# Overtonium, the whole site

> A free, open source additive synthesizer plugin laid out like a 32-channel mixer. Every channel is one sine oscillator locked to a harmonic of the played note, and every channel has its own tuning, envelope, modulation and place in the stereo field. VST3, Audio Unit, LV2 and standalone, for macOS, Windows and Linux, under the AGPLv3.

Every page of https://benjamindehli.github.io/overtonium/ as Markdown, in the order the site lists them, so the whole of it can be read in one request. The short version, with links into each page, is https://benjamindehli.github.io/overtonium/llms.txt.

---

Page: Overtonium, a free 32-partial additive synthesizer plugin
URL: https://benjamindehli.github.io/overtonium/
Last changed: 2026-11-07

# Overtonium, a 32-partial additive overtone synthesizer

32-partial overtone synthesizer

An additive synthesizer laid out like a 32-channel mixer. Every channel is one sine oscillator locked to a harmonic of the played note, and every channel has its own tuning, envelope, modulation and place in the stereo field.

- [Download](https://benjamindehli.github.io/overtonium/install/#downloads)
- [Source on GitHub](https://github.com/benjamindehli/overtonium)
- [Star the repo](https://github.com/login?return_to=%2Fbenjamindehli%2Fovertonium)
- [Follow @benjamindehli](https://github.com/login?return_to=%2Fbenjamindehli)

**Formats**

VST3, Standalone, AU on macOS, LV2 on Linux

**Platforms**

macOS, Windows, Linux

**Partials**

32, plus a noise channel

**Licence**

AGPLv3

Additive synthesis usually hides its partials behind a spectrum drawing or a few macro controls. Overtonium puts all of them on the surface as faders, and gives each one the controls of a channel on a mixing desk. You shape a sound by mixing it.

[Image: The Overtonium window: a top bar of global controls above 32 channel strips and a noise channel, with the meters lit by a held chord.]

The whole instrument. Thirty-two channels, a noise channel pinned on the right, and the global controls along the top.

## What it sounds like

Eleven of the forty-four [factory presets](https://benjamindehli.github.io/overtonium/presets/), played on an Expressive E Osmose, with nothing on the output except the plugin's own effects. All the other audio on this site is the plugin rendering a chord to a file. That keeps those clips accurate, but it cannot show how the plugin responds to playing. This is the only recording of it being played.

[Video: Overtonium sound examples](https://www.youtube.com/watch?v=L1oYdPxGlGA)

[The presets page](https://benjamindehli.github.io/overtonium/presets/#video) has a chapter link for each of the eleven, and a clip of every one of the forty-four presets.

## The control to try first

TUNE sweeps each partial continuously between equal temperament and just intonation. At the just end, each partial sits at an exact whole-number ratio to the fundamental and the whole stack blends into one timbre. At the equal end, each partial snaps to the nearest semitone and the same stack sounds like a chord.

Two of the [factory presets](https://benjamindehli.github.io/overtonium/presets/), **Just Saw** and **Equal Saw**, are identical except for that one control. They sound completely different.

**Just Saw**

**Equal Saw**

One note each, the same patch, the same level. Only TUNE differs.

Nothing about the tuning is hard-coded. For harmonic `n` the just interval is `1200 * log2(n)` cents, the equal position is that rounded to the nearest semitone, and TUNE dials in the remainder. [The tuning system](https://benjamindehli.github.io/overtonium/tuning/) explains it in detail, along with stretch, the nine keyboard temperaments and keyboard tracking.

## Where to go next

- [**Tuning**TUNE, inharmonic stretch, eight historical temperaments and an adaptive one on any root, and how the spectrum thins out higher up the keyboard.](https://benjamindehli.github.io/overtonium/tuning/)
- [**Controls**Every knob on a channel strip, then the effects on the bar and how to play it from a keyboard, a host or a controller.](https://benjamindehli.github.io/overtonium/controls/)
- [**Presets**The forty-four that ship, what a preset carries, and where your own are kept on each platform.](https://benjamindehli.github.io/overtonium/presets/)
- [**Install**Installers for macOS and Windows, a zip for Linux, what to do when a host does not see it, and building from source.](https://benjamindehli.github.io/overtonium/install/)

What changed in each version, newest first, and what to know before upgrading are on [the releases page](https://benjamindehli.github.io/overtonium/releases/).

## Questions

**Which hosts does Overtonium run in?**

It comes as a VST3 on macOS, Windows and Linux, an Audio Unit on macOS and an LV2 on Linux, so it loads in any host that supports those formats. There is also a standalone build.

**Is Overtonium free?**

Yes. It is also free software, not only free of charge. It is released under the AGPLv3 because it uses JUCE, and all of the source is on GitHub.

**What is additive synthesis?**

It builds a sound by adding sine waves together, instead of filtering a rich waveform. Every timbre is a stack of partials at different frequencies and levels, and additive synthesis lets you control those partials directly. Overtonium gives you 32 of them, one per mixer channel.

**Can Overtonium do microtonal tunings and just intonation?**

Yes, both, and they are set separately. TUNE moves each partial between its equal-tempered position and its exact whole-number ratio. The keyboard has its own tuning: equal, just, Pythagorean, quarter-comma meantone, Werckmeister III, Vallotti, Kirnberger III or Young, on any root, at any of eleven reference pitches from 415 to 466 Hz. It can also use [Adaptive](https://benjamindehli.github.io/overtonium/tuning/#adaptive), which keeps the chord you hold in just intonation as you play, and moves a held note when the next chord needs it.

**Does Overtonium support MPE?**

Yes. With [MPE](https://benjamindehli.github.io/overtonium/playing/#mpe) switched on, pitch bend and pressure work per note, and each of the 33 channels has its own pressure amount. An ordinary single-channel keyboard works either way.

**Can I play Overtonium from a MIDI controller?**

Yes. Right-click any control, choose [MIDI Learn](https://benjamindehli.github.io/overtonium/playing/#midi-learn) and move a knob or fader on the controller. The mapping is saved with your project, not with the preset. A [macro](https://benjamindehli.github.io/overtonium/playing/#macros) is one parameter that moves a whole row, so one controller or one automation lane can move up to 32 controls at once.

**Does Overtonium have portamento or glide?**

Yes, on each channel separately. Each partial slides to a new note over its own GLIDE time, so the partials can arrive one at a time. Two switches apply to every channel. One sets whether it glides always or only when you play legato, and the other whether it moves at a fixed rate or takes a fixed time. [Glide](https://benjamindehli.github.io/overtonium/playing/#glide) on the Playing page has the details.

**How much CPU does Overtonium use?**

Eight voices of 32 partials each, with every modulator running, use about 7% of one core, and sixteen voices about 14%. The effects add little, except the resonator with all its strings free, which costs about half as much as eight voices. Lowering the render rate saves a lot. At 8 kHz the whole voice pool costs a fifth as much. [The full figures](https://benjamindehli.github.io/overtonium/install/#cost) are on the install page.

## Where it came from

Overtonium is the fourth instrument in a series, and the first that computes its sound instead of playing back recordings. The two ideas it is built on both came from sitting at a drawbar organ, and neither is really about organs.

The first was the [Voltage Controlled Cassette Organ](https://github.com/benjamindehli/VoltageControlledCassetteOrgan), in 2023. It is a Korg CX-3 sampled through cassette tape, with every note of every drawbar recorded as its own file. The tape was the main idea: wow, flutter and random warbles gave the organ a movement like a chorus or a rotary speaker, but without the fixed rate of a modulation effect. It was recorded to tape twice, which doubled it, and because no tape deck plays back exactly the same way twice, the two passes never quite matched.

[Strykebrett](https://github.com/benjamindehli/Strykebrett) followed in October of the same year, a sampled Logan String Melody II. It has six faders instead of nine drawbars, each setting the level of one octave or one of the voices across a bass and a treble section, and four of them have their own envelope and their own LFO. It is not an organ, and that is what makes it useful here: the workflow still worked on a different instrument.

[EDB-Orgel](https://github.com/benjamindehli/EDB-Orgel) followed in 2024. It keeps the drawbar workflow but puts four digital synthesis types under it instead of an organ.

All three gave each drawbar its own envelope and LFO, which is the first of the two ideas. A drawbar is a partial, and on an organ you move the drawbars by hand while the note sounds. Nine of them shaped separately already sound unlike an organ. Thirty-two, each with a full channel of controls, take that idea as far as it goes.

The second idea is about how the drawbars are tuned. A drawbar is a harmonic, but an organ only has the twelve notes of the keyboard to build it from, so each drawbar is tuned to the nearest one. The second, fifth and eighth are all fifths an octave apart, so holding a C puts three G's into the sound. A harmonic series does not work like that. Its real intervals are whole-number ratios, 2:1, 3:2, 4:3, and apart from the octaves none of them lands on a key. [TUNE](https://benjamindehli.github.io/overtonium/tuning/) turns that difference into a control, and computing the partials is what makes it possible. A sampled partial is stuck at the pitch it was recorded at. A computed one can sit where the organ would put it, at its exact ratio, or anywhere between.

You can hear the ancestry in the details. DRIFT is the cassette's random warble, per partial. WOBBLE is the same thing applied to the whole instrument. The tape echo's two paths, each with its own motor speed and its own random stream, are the double tracking. Its AGE control never quite reaches zero for the same reason the two tape passes never matched: a transport that held speed exactly would put the repeat back in mono, and no real transport holds speed exactly.

All three earlier instruments are sample libraries with a plugin wrapper, sold at [Dehli Musikk](https://store.dehlimusikk.no/). This one is free software, and none of their code is in it.

---

Page: The Overtonium tuning system: TUNE, stretch and nine temperaments
URL: https://benjamindehli.github.io/overtonium/tuning/
Last changed: 2026-11-06

# The Overtonium tuning system

Three separate controls. TUNE sets how the overtone series is tuned, STRETCH sets how far it moves away from a true harmonic series, and the temperament sets how the keyboard itself is tuned. They work together and do not overlap.

## TUNE, from equal to just

A drawbar organ is the easiest way to explain this control, because it solves the same problem the other way round. Each drawbar is a harmonic, but an organ only has the twelve notes of the keyboard to build it from, so each drawbar is tuned to the nearest one. The second, fifth and eighth are all fifths an octave apart, so holding a C puts three G's into the sound. A harmonic series is not spaced like that. Its intervals are whole-number ratios, 2:1, 3:2, 4:3, and apart from the octaves none of them lands on a key. TUNE moves between those two tunings. It is a continuous knob, not a switch.

For harmonic `n`, the just interval above the fundamental is `1200 * log2(n)` cents. Overtonium rounds that to the nearest semitone to get the equal-tempered position. The remainder is the cent offset that the TUNE knob adds.

```
semitoneOffset(n, blend) = round(1200*log2(n)/100) + blend * (1200*log2(n) mod 100) / 100
```

At `blend = 1` this is exactly `n` times the fundamental, so the stack blends into one timbre. At `blend = 0` every partial is on a semitone of the chromatic scale and the same stack sounds like a chord. None of the values are stored as constants. Rounded to whole cents, the formula gives the familiar table, and the test suite checks this so it cannot change by accident.

The same patch on the same note, with only TUNE changed. First every partial at its exact whole-number ratio, then each one moved to the nearest semitone, then the knob turned slowly from one to the other.

**Just Saw**

**Equal Saw**

**The whole sweep**

[Diagram: Where each partial sits against the nearest semitone. Thirty-two rows, one per partial. The line down the middle is the tempered semitone that TUNE moves the partial to at one end of its range. Each dot is where that partial falls as a whole-number ratio. The knob moves the partial across the distance between them. Only the octaves, partials 1, 2, 4, 8, 16 and 32, sit exactly on the line. The table below gives the exact figures. -50 -25 the semitone +25 +50 n cents from the nearest semitone 1 2 3 4 5 6 7 -31 8 9 10 11 -49 12 13 +41 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32.]

Every partial compared with the semitone TUNE moves it to. The dot is the whole-number ratio and the line is the tempered position. The knob moves the partial across the distance between them. Only the octaves sit on the line.

This means that on channels 1, 2, 4, 8, 16 and 32 the knob does not change the sound. An octave is 1200 cents in both tunings, so the blend has nothing to move on those six, and their readouts stay at 0.0 wherever the knob is. Hovering over one of those strips tells you this. The channels are not stuck. Only that knob leaves them alone. STRETCH moves them, and so do DRIFT, the pitch modulator and the oscillator character, whose rack puts every partial except the first a few cents off its exact pitch.

| n | Semitones | Cents | Interval | n | Semitones | Cents | Interval |
|---|---|---|---|---|---|---|---|
| 1 | 0 | 0 | prime/octave | 17 | 49 | +5 | minor second |
| 2 | 12 | 0 | prime/octave | 18 | 50 | +4 | major second |
| 3 | 19 | +2 | fifth | 19 | 51 | -2 | minor third |
| 4 | 24 | 0 | prime/octave | 20 | 52 | -14 | major third |
| 5 | 28 | -14 | major third | 21 | 53 | -29 | fourth |
| 6 | 31 | +2 | fifth | 22 | 54 | -49 | tritone |
| 7 | 34 | -31 | minor seventh | 23 | 54 | +28 | tritone |
| 8 | 36 | 0 | prime/octave | 24 | 55 | +2 | fifth |
| 9 | 38 | +4 | major second | 25 | 56 | -27 | minor sixth |
| 10 | 40 | -14 | major third | 26 | 56 | +41 | minor sixth |
| 11 | 42 | -49 | tritone | 27 | 57 | +6 | major sixth |
| 12 | 43 | +2 | fifth | 28 | 58 | -31 | minor seventh |
| 13 | 44 | +41 | minor sixth | 29 | 58 | +30 | minor seventh |
| 14 | 46 | -31 | minor seventh | 30 | 59 | -12 | major seventh |
| 15 | 47 | -12 | major seventh | 31 | 59 | +45 | major seventh |
| 16 | 48 | 0 | prime/octave | 32 | 60 | 0 | prime/octave |

The two large offsets are the ones you hear most. The seventh partial is 31 cents flat of a minor seventh, and the eleventh is 49 cents flat of a tritone, about halfway between two keys. These partials make a just-tuned stack sound unlike anything a keyboard can play.

In Norway the eleventh has a name. Its pitch, between the fourth and the tritone, is the natural fourth of Norwegian folk music. The willow flute plays it because it can only play the harmonic series. The composer Eivind Groven spent much of his life on it. From 1936 he built organs with up to three pitches for every key, chosen automatically as the music moved. This let an ordinary keyboard play pure triads in every key, with the eleventh partial exactly in tune. In Overtonium it is the eleventh row with TUNE at just, 551 cents above the octave below it. His way of keeping every chord pure on an ordinary keyboard is the [Adaptive](https://benjamindehli.github.io/overtonium/tuning/#adaptive) temperament further down.

## Stretch

Real instruments do not ring at exact whole-number multiples. A string with any stiffness has partials at `n * f0 * sqrt(1 + B * n^2)`. They are sharp of the harmonic, and more so further up the series. This, more than the hammer or the soundboard, is what makes a piano sound like a piano and not like a sawtooth. It is also why a tuner stretches the octaves. If the top of the instrument is tuned to the theory, it beats against its own overtones.

STRETCH sets that stiffness. It is set by how far it moves the top partial, not by the stiffness value itself, which runs from 0.00003 to 0.008 and is hard to read. At +150 cents, which is a real piano:

| Partial | 2 | 8 | 16 | 32 |
|---|---|---|---|---|
| Cents sharp of harmonic | +0.6 | +10.2 | +40.0 | +150.0 |

The bottom of the series barely moves and the top moves a long way. Turn it further and the partials no longer share a fundamental, so the sound stops being a note with a timbre and becomes a bell. Negative values pull the partials inward. No real string does this, but it is still a useful sound.

It is separate from TUNE, and the two work together. Equal temperament with heavy stretch sounds different from just intonation with the same stretch. Zero is the plain harmonic series.

The [resonator](https://benjamindehli.github.io/overtonium/effects/#resonator)'s strings are just as stiff, so a stretched partial still finds the string it should ring. Up to about a piano's worth of stretch, the first sixteen overtones of each string land within a fraction of a cent of the partials. Further out the upper ones drift. The low strings can only be pulled in a little, so with negative stretch it is mostly the high strings that follow.

One note while STRETCH turns from a harmonic series to beyond a piano. The bottom of the series barely moves and the top moves a long way. Along the way the partials stop sharing a fundamental.

**STRETCH from 0 to 500 cents**

## Tuning the keyboard

Everything above is about where a partial sits above the note you play. This section is about how that note is tuned. Settings and the [Tuning bar](https://benjamindehli.github.io/overtonium/playing/#bars) have nine temperaments: eight fixed ones, and one that retunes as you play.

| Temperament | Major third | Fifth | What it is |
|---|---|---|---|
| Equal | 400.00 | 700.00 | the reference everything else is measured against |
| Just (major) | 386.31 | 701.96 | pure by construction, chosen interval by interval |
| Pythagorean | 407.82 | 701.96 | every fifth pure, at the cost of a wide third |
| Quarter-comma meantone | 386.31 | 696.58 | fifths narrowed a quarter comma, which makes the third pure |
| Werckmeister III | 390.22 | 696.09 | four fifths narrowed, no wolf, every key playable |
| Vallotti | 392.18 | 698.04 | six fifths, F-C to E-B, narrowed by a sixth of a comma, milder still |
| Kirnberger III | 386.31 | 696.58 | four fifths narrowed a quarter comma, so the third on the root is pure, no wolf |
| Young | 392.18 | 698.04 | Vallotti's six fifths a fifth higher, C-G to B-F# |
| [Adaptive](https://benjamindehli.github.io/overtonium/tuning/#adaptive) | 386.31 | 701.71 | pure in every key, the held chord retuned as you play |

They are worked out from the circle of fifths, not copied from tables of cents. Almost every historical temperament is described by how much each of the twelve fifths is narrowed. The code stores that, and the pitch classes are calculated from it. The two commas everything is built from come out at 23.460 and 21.506 cents, which match the published figures. The tests check the property that defines each temperament, not the resulting numbers: meantone's third pure to a thousandth of a cent, Pythagorean's fifth the same, Werckmeister's third at 390.2, and Kirnberger III's third pure.

Vallotti and Young narrow the same six fifths, one starting a fifth lower than the other. So Vallotti on C is Young on F, note for note. Both are in the list because players know them by both names.

### Adaptive

Every fixed temperament is in tune in some keys and out of tune in others. Just major is pure in the keys near its root, and a chord in a key further away is a comma out. This is the oldest problem in tuning. Adaptive solves it the way Eivind Groven's organs did. It gives every key more than one pitch and chooses between them as you play. Every pitch class has five versions: its place in just major on the root, and one and two commas either side of it. The chord you hold uses whichever versions make it purest.

The chord is retuned every time a key goes down or comes up. A chord that changes into the next one a note at a time stays in tune all the way, as well as a chord struck all at once. Keys struck together are tuned together. If the new chord needs a held note at a different pitch, the note slides there over the time set by the **Retune** knob in the Tuning bar, from an instant jump at zero up to a second. At 30 ms, where the knob starts, the note sounds as if it was always there, and at around 300 ms you can hear the chord settle. Settings offers those two as *Fast* and *Smooth*. Adaptive moves as few notes as it can, and new notes take whichever version fits the notes already held. The keys you hold decide how the chord is tuned, not the order you pressed them in. A chord struck at once, played a key at a time or reached from another chord comes out the same, interval for interval. Every chord type in [music-utils](https://github.com/benjamindehli/music-utils), my library for naming chords from the notes held, is played through it in all twelve keys on every build. That is 125 of them, from a bare third to an altered thirteenth. The test checks that each chord comes out the same every way, and that it is tuned as purely as it can be.

It tunes against everything you can hear: the keys that are down and the ones the pedal is holding. When the held keys allow more than one version, a chord still ringing after its keys came up decides between them. A key struck into silence takes the version its pitch class used last. This keeps a progression from wandering away a comma at a time. No key can go further than its outer versions, so the keyboard never drifts more than two commas from where it started, however long you play. If every chord were tuned pure to the last one with no limit, a C major, A minor, D minor, G progression played four times would end up over two and a half semitones sharp. If you want no drift at all, turn on the **NO DRIFT** switch in the Tuning bar, or No drift in Settings. It keeps the chord as close to home as it can. When a progression would carry it away, a held note slides a comma instead, and a key struck into silence goes back home.

Like Groven's organs, it narrows each fifth by an eighth of a schisma, a quarter of a cent, so that eight fifths and a pure major third meet exactly. Without this, chords far from the root would be two cents out, which on pure partials gives slow but clearly audible beating. The major thirds are pure. A quarter of a cent on a fifth beats about once every eighteen seconds on the third partial of a note in the middle of the keyboard. The same chord reached by a different route can end up a comma higher or lower, with every interval in it the same, so a piece can end a comma or two from where it began. Root sets where the middle versions sit, and Reference pitch sets where A's middle version sits.

Some chords cannot be pure all the way through, because pure fifths and pure thirds never quite fit together. In C11 the fifth from G to D and the minor third from D to F cannot both be pure. Adaptive then gives up as little as it can. It ranks a fifth or a fourth above a third or a sixth, and those above a seventh. Counted that way, C6 with an added ninth keeps its thirds and sixths and lets the fourth from A to D go a comma wide. Keeping that fourth would cost three thirds.

A minor seventh is tuned to the harmonic seventh, 7:4 above its root, 969 cents, unless that would put it out of tune with another held note. A dominant seventh becomes 4:5:6:7. These are four harmonics of one note an octave below its root, the same seventh that every partial here already has as its seventh harmonic, and the smoothest a seventh chord can be. It is noticeably lower than the equal-tempered seventh, and that is most of why it sounds so settled. A ninth adds the ninth harmonic. A seventh over its root alone or over its fifth is 7:4 as well. With a minor third above the root, the seventh stays at 9:5, because 7:4 would put it 48 cents flat of a pure fifth above that third. A sharp ninth is held back the same way, for the same reason. A suspended seventh keeps 16:9, two pure fourths. With a minor second or a tritone above the root, the seventh also keeps the size the chain gives it. A major seventh is 15:8. Like everything else here, it depends only on the keys held. C and B flat are 7:4 however they were played, and a held note slides to or away from the harmonic seventh when the chord around it changes type. The **7:4** switch in the Tuning bar, and Harmonic sevenths in Settings, turn this off. Every seventh then takes the size the chain gives it.

The [Tuning bar](https://benjamindehli.github.io/overtonium/playing/#bars) shows how Adaptive is tuning the chord you hold, and updates as each key goes down and comes up. **PURE** lights while every interval in the chord is pure. It goes dark when a chord cannot be pure all the way through and one interval has been given up. **7:4** lights while a seventh is tuned to the harmonic seventh, and **DRIFT** shows how far the keyboard has drifted, in commas. Next to those, there is one display for each of the twelve notes, showing how far it is from equal temperament in cents. The notes you hold are lit, and the rest stay unlit until you play them. Under every other temperament the same twelve displays show the whole scale, all of them lit. So the Tuning bar shows what Werckmeister III or Just on D does to each note before you play it.

[Image: The Tuning bar under Adaptive with C7 held: the temperament reading Adaptive on a root of C at 440 Hz with a drift of 0, NO DRIFT off, 7:4 lit, PURE lit and the Retune knob near its start, then twelve displays for the notes, C, E, G and B flat lit at 15.4, 1.7, 17.1 and minus 15.8 cents, the white keys on a light ground.]

C7 under Adaptive. C is 15.4 cents above equal, where just major on C puts it with A at 440. E and G are a pure third and fifth above it, and B flat is the harmonic seventh, 31.2 cents below C's equal-tempered seventh.

### Root

Sets which pitch class the temperament is built on. An unequal temperament is only in tune in the keys near its centre, so you choose that centre here. For Adaptive it sets where the middle versions are placed. It is greyed out for equal temperament, which has no centre, unless the [resonator](https://benjamindehli.github.io/overtonium/effects/#resonator) is in Series or Fifths, which are built on the root in every temperament.

On a root of C every temperament is laid out the way organs were tuned. Pythagorean and quarter-comma meantone each have one fifth that cannot be used, the wolf, and on C it is G#-Eb. That keeps C, F, Bb, G, D and A major in tune. On another root the wolf moves with it, a minor sixth above the root.

### Reference pitch

One of eleven pitches for A: 415, 430, 432, 435, 438, 440, 442, 443, 444, 446 or 466 Hz. A stays exactly at this pitch in every fixed temperament, because the offsets are measured relative to A, the same way a tuner tunes A first and works outwards. Only the other eleven notes move. Under Adaptive, A's middle version sits on the reference, and an A in a chord that needs it takes the version a comma either side.

The [resonator](https://benjamindehli.github.io/overtonium/effects/#resonator)'s strings are tuned the same way, so they follow the temperament, the root and the reference pitch. Its Series mode is built on the root's own harmonics, and its Fifths mode on the root and its fifth.

None of the three is saved with a [preset](https://benjamindehli.github.io/overtonium/presets/). A temperament belongs to the music you are playing, not to any one sound in it. You set it once and keep working, and a patch that reset it to equal while you worked would get in the way. This includes Init, which clears the patch but not the session.

One C major chord, twice, on a temperament built on C. Equal puts the third at 400 cents and Werckmeister III at 390, which is most of the way to the pure 386. The beating between the third and the root slows almost to a stop.

**Equal**

**Werckmeister III**

## Tracking

Without TRACK, a patch has the same spectrum at every pitch. Real instruments do not. Play up a piano and the top of the series thins out. This happens because the body has a rolloff at a fixed frequency, and the partials climb up through it as the notes get higher.

TRACK sets that rolloff, in dB per octave above a fixed corner of 1 kHz, which is around C6. It is measured relative to the fundamental, not as an absolute level. An absolute rolloff would act as a shelf and make high notes quieter instead of duller. Measured this way, the fundamental keeps its level at every pitch and only the spectrum above it thins.

The effect depends on where the corner sits. A bass note has most of its series below 1 kHz and keeps nearly all of it. A treble note whose fundamental is already above the corner gets the full slope on every partial. At 6 dB per octave, the 32nd partial keeps 57% of its level at A1 and 4% at A5. Zero is off, and off changes nothing.

The same five notes an octave apart, twice. With TRACK off the spectrum is the same at every pitch. With it on, the bass note keeps nearly all of its series, because most of it sits below the corner, and most of the top note's series is in the rolloff. Measured on these two files, the partials above the fundamental add up to 1.43 times the fundamental at C2 in both. At C6 they add up to 1.45 with tracking off and 0.49 with it on.

**TRACK off**

**TRACK at 6 dB/oct**

## Drift, the fourth kind of detuning

DRIFT works per channel, not globally, and it is random, not periodic. It moves a partial's pitch in a smooth random path of up to 25 cents that keeps changing as the note plays. This is the cassette warble that the earlier instruments in this line got from tape. Every partial has its own random stream, so a stack with drift never quite settles, and no two notes drift the same way.

Drift is why a held chord keeps moving with no modulation set. It is also why the pitch lamp on the PITCH MOD rule shows a needle and not a pulse. The needle shows where modulation and drift together put the partial at that moment.

One held chord, twice. With drift off, each partial stays exactly at its pitch. With it on, every partial moves on its own and the stack never settles. Measured across these files, the eighth partial stays at 0 cents in the first and moves over 12 cents in the second.

**DRIFT off**

**DRIFT at 18 cents**

---

Page: Overtonium controls: every knob on a channel strip
URL: https://benjamindehli.github.io/overtonium/controls/
Last changed: 2026-11-04

# Overtonium controls

Thirty-two channel strips, a noise channel, and a bar of global controls across the top. Every strip has the same twenty-four controls. That is seven hundred and sixty-eight of them before the bar is counted, so much of the design is about not having to move them one at a time.

This page covers the mixer. The bar across the top, from the oscillator character to the output stage, is covered on [Effects](https://benjamindehli.github.io/overtonium/effects/). Playing the instrument from a keyboard, a host or a controller is covered on [Playing](https://benjamindehli.github.io/overtonium/playing/).

## A channel strip

[Image: A close view of the first six channels, showing the row captions in the gutter with the pointer tool button above them, and the interval colour of each channel.]

Six channels and the caption gutter. Each row is named once on the left instead of on every strip.

Hold the pointer over a row's name in the gutter to see what the row does. The table below says the same at more length.

| Control | Range | What it does |
|---|---|---|
| TUNE | equal to just | Where this partial sits between the tempered position and the exact ratio. The readout shows the resulting cent offset. On the six octave channels both tunings agree, so the knob has nothing to move and the readout stays at 0.0 |
| PITCH MOD shape, rate, depth | Eight shapes, 0.01 to 30 Hz, 0 to 1200 cents | Per-partial vibrato |
| DRIFT | 0 to 25 cents | Smooth random pitch wander, its own stream per partial |
| GLIDE | Off to 5 s | How long this partial takes to reach a new note, so the partials can arrive one at a time. The L and T switches beside the caption, and the knob's right-click menu, set whether it glides always or only legato, and at a fixed rate or in a fixed time. See [Glide](https://benjamindehli.github.io/overtonium/playing/#glide) |
| ENVELOPE strike, delay, A, D, S | -100 to +100%, 0 to 5 s, 90° then 0.2 ms to 5 s, 1 ms to 20 s, 0 to 100% | Exponential decay. The harder the key is struck, the sooner and faster STRIKE starts the partial |
| KEY OFF swell, level, release | 0 to 5 s, 0 to 100%, 1 ms to 20 s | A second envelope that runs when the key is released |
| AMP MOD shape, rate, depth | Seven shapes, 0.01 to 30 Hz, 0 to 100% | Per-partial tremolo |
| VELOCITY | -100 to +100% | How much key velocity scales this partial. Negative inverts it. The dB switch beside the caption picks the curve |
| KEYBOARD | -200 to +200% | How the level follows the key played. Up to 100% it tilts around middle C, and past 100% the turn moves away from middle C |
| AFTERTOUCH | -100 to +100% | How much key pressure moves this partial. Negative fades it out |
| PAN | hard left to hard right | Equal power, so the level stays the same as the partial moves across the stereo field |
| M and S |  | Mute overrides solo. Option-click or Alt-click either one to mute or solo only that channel. Right-click either one for the same, or to clear every mute or every solo at once |
| LEVEL | -inf to 0 dB | Fader scaled in decibels, with the meter inside its track |

## What the modulators trace

Both per-partial modulators have a shape as well as a rate and a depth. You pick the shape from a small drawing of the wave between the two knobs. Click it to see the list by name. Pitch has eight shapes and amplitude has seven. Each channel has its own, the noise channel included, so the fifth partial can switch on and off while the fourth swells smoothly.

| Shape | What it does |
|---|---|
| Sine | The default |
| Triangle | The same range at an even speed, which sounds more mechanical than a sine |
| Sawtooth, Reverse Sawtooth | A ramp and a sudden jump back, in either direction |
| Square | Two values and nothing in between. On pitch it is a trill. It comes in a bipolar form, which goes the same distance above and below the note, and a unipolar form, which only goes up |
| Sample & Hold | A new random value each cycle, held until the next. On pitch, with the depth at full, it plays random notes anywhere within an octave above or below |
| Random | The same random values, moving smoothly between them instead of in steps. It is the contour DRIFT follows, at a rate you set |

**Amplitude has only one square.** A tremolo only ever lowers the level set by the fader, so a unipolar shape has no second direction to go in. The one square switches across the whole depth. For the same reason, every amplitude shape starts a quarter of a cycle ahead. A note begins at full level and then dips, instead of starting half attenuated.

The square and stepped shapes do not click. A partial's gain changes smoothly across each control block instead of jumping, so each edge takes about two thirds of a millisecond. That is short enough to sound like an edge and long enough not to be heard as a click.

LINK does not change the shapes. It moves a value across the series along a weighted curve, and a list of named shapes cannot be weighted. Instead, the shape menu can set every channel at once. **Set every channel to this** copies the channel's own shape to all the others, and **Set every channel to** lists every shape, so all the channels can change to a new shape in one step without changing one of them first. Either one is a single step to undo.

**Shared by all notes** is the other entry in that menu, one for each of the two modulators. It is also the switch beside each SHAPE caption on the left of the mixer, marked with three strokes joining into one. By default each note has its own modulator. Play a chord one note at a time and you get three tremolos, each starting when its own key went down, so they move out of step. With the entry ticked, the modulator belongs to the channel instead. A note that arrives late joins the modulation already running, so the chord moves as one however it was played. An instrument with a single tremolo circuit behaves like this, which is why [Wurli](https://benjamindehli.github.io/overtonium/presets/), Tape Pluck and Soft Trio ship with their tremolo shared, and StyloPoly and Combo Organ with their vibrato.

Each channel keeps its own rate and shape either way. Sharing gives one modulator per channel, not one for the whole instrument, so a patch built on thirty-two rates that never line up still sounds the same with the switch on. The switch is part of the patch, so a preset saves it. Both switches are off unless a patch turns them on.

The switch lights when its modulator is shared, so a preset with a shared tremolo shows it on the panel as well as in the menu. There is one switch per modulator, and it applies to all thirty-three channels.

## Velocity and pressure, per partial

Because velocity is set per partial, a soft note can have a different timbre, not just a lower level. Set the fundamental to 0% and the upper partials to 100%, and the tone gets brighter as you play harder. A struck string or bar behaves roughly like this.

Both controls go either side of zero. A positive amount makes harder or heavier playing louder. A negative amount inverts that, so the partial is loudest when you play softly. The two halves mirror each other exactly, so -50% at a given velocity matches +50% at the opposite velocity. The negative half is useful for crossfading. Give one set of partials a positive amount and another set a negative one, and the two timbres swap places across the velocity range instead of one just fading in.

**The dB switch** beside the VELOCITY caption picks how velocity becomes level, for every channel at once. Lit, each step of velocity takes the same number of decibels off. At 100% a medium note is 20 dB down and the softest note 40 dB down, and at 50% each of those is half as much. Dark is the Classic curve, which takes velocity off as a fraction of the level. Even at 100% it leaves a medium note only 6 dB down, and most of what it does happens on the softest notes, so it only starts to sound like much near the top of the knob.

Patches made before the switch existed keep Classic, so a saved project and a saved preset sound as they did. The factory presets use dB. Clicking the switch converts every channel's amount with it, so a medium note keeps its level and only the soft and loud ends move. Classic at 100% becomes about 30% on dB. Both the switch and the conversion are a single step to undo.

Aftertouch works the same way but **adds** to the fader instead of scaling it, and it ignores velocity. A strip with its fader all the way down is silent until you press into the key, and then it fades in. A negative amount fades an open strip out instead. Both channel pressure and polyphonic aftertouch are accepted, and the higher of the two is used. Pressure is smoothed over about 15 ms, so the steps of seven-bit MIDI are not heard in the gain.

By default the mod wheel also works as aftertouch, because most keyboards have no aftertouch and the wheel is close to hand. A wheel left at rest reads zero, so nothing changes on a controller that does send pressure. To use only one or the other, choose it in the "Aftertouch from" entry in Settings.

## Following the keyboard, per partial

KEYBOARD turns a partial down depending on which key is played, so a patch can be voiced across the keyboard the way an organ builder voices a stop: a fundamental that is weaker in the bass, bell partials only on the top keys, or the noise only in one range. It only ever turns a key down, so a partial is never louder than its fader. At zero, the default, it does nothing.

Up to 100% either way it tilts around middle C, which it never changes. A positive amount turns down the keys below middle C and a negative one the keys above it. At 100% the slope is 24 dB per octave. The knob is shaped so the gentle slopes get most of its travel: 50% is 6 dB per octave and 25% is 1.5.

**Past 100% the slope stays at 24 dB per octave and the turn moves** away from middle C, an octave for every 20%. A positive amount moves it up, so only the treble is left, and at 200% it is at C9, near the top of the MIDI range. A negative amount moves it down, and at -200% it is at C-1, the bottom. Turning the knob further from zero always turns more of the keyboard down, so a macro, LINK or automation moving it does more of the same all the way. A moved turn always comes with the steepest slope, which makes a near split, with a partial fading out over a few keys.

The readout under the knob gives the slope and the key the fade starts from, such as "6.0 dB/oct below C4" or "24 dB/oct above F#1". Middle C is C4.

It reads the key number, not the pitch, so pitch bend, TUNE, the temperament and drift leave the level alone. When a glide carries a note to a new key, the level follows the note as it slides. It also scales what AFTERTOUCH adds, so pressure cannot bring a partial back on keys the row has faded it out of. TRACK in the bar does a different job: it dulls the top of every partial alike, by frequency. KEYBOARD is set per channel and can go either way.

## The two-part envelope

Delay, attack, decay and sustain while the key is down. Releasing the key runs a small envelope of its own instead of just fading out. The KEY OFF rows set the level the sound moves to when the key is released, and how long it takes to get there. Only then does the release run from that level down to silence.

A key-off level above the sustain gives a release click or swell, like the sound of a damper landing or a hammer returning, and it can be louder than the held note. *Drawbar Organ* uses it this way: the upper drawbars jump to 55% for two milliseconds as the contacts break. A level below the sustain gives the fast drop into a long tail that a piano or a struck bell has, which is probably the more common use. A key-off level of zero skips the stage and releases from the current level.

**How fast you release the key scales that tail**, on a keyboard that can sense release velocity. The fastest release doubles the level, the slowest halves it, and 64 leaves it unchanged. It scales the level the tail starts from. On a patch with its own key-off level it changes that level, and on a patch without one it changes the sustain. It has no knob and no parameter, because like velocity it belongs to how you play, not to the patch. A fast release cannot raise a partial that is already at full level, since the envelope stops at full scale.

Most keyboards cannot sense release velocity, and on those nothing changes. They send either a note-off with a velocity of zero or a note-on with a velocity of zero in its place. Both are read as no information, not as the slowest possible release.

The delay stage keeps a partial silent before its attack begins. Setting different delays across the series makes the spectrum build up gradually instead of arriving all at once. This is how *Slow Pad* and *Shimmer* open up. The delay is fixed in samples at note-on, so moving the knob does not change the timing of a note already waiting.

**ATTACK goes below its shortest time.** Turned below 0.2 ms, it sets where in its cycle the partial starts instead of a time. This runs from 0°, a rising zero crossing, to 90° at the bottom of the knob, which starts the partial at its peak with no attack at all. A zero crossing is the softest possible start. A partial cannot reach its peak until a quarter of its period has passed, and below about 500 Hz that is longer than the shortest attack. So on most of the keyboard, where the cycle starts shapes the start of a note more than the envelope does. One millisecond into an A1, a partial starting from the zero crossing is a third of the way up, one starting from 45° is nine tenths of the way up, and one starting from the peak is all the way up. This part of the knob's ring is marked with dots instead of ticks, so you can see at a glance whether a knob is set in degrees or in milliseconds.

Like every row, it is set per channel, so some partials can start sharply and others not. LINK can spread it across the series in one drag. If every partial starts at its peak at the same time, the result is a spike: for 32 partials on a falling series, the first sample is about four times the steady peak, and the output stage has to handle it. Spreading the start points across the channels keeps the sharp start without the spike. With phase reset off in Settings, a partial carries on from wherever its cycle was, and this part of the knob acts as the shortest attack. The noise channel has no cycle to start, so its ATTACK stops at 0.2 ms.

**STRIKE** is at the top of the section because it is about how hard the key is struck, not about the shape of the envelope. It applies velocity to the two rows below it, the way the velocity row applies it to the fader. At zero it has no effect. A positive amount makes a hard note start sooner and rise faster, and makes a soft note wait and then rise slowly. A negative amount inverts this.

The two act in opposite directions on purpose, so each knob stays the limit of its own row. Velocity only ever lengthens the attack, so ATTACK is always the fastest the partial can rise. It only ever shortens the delay, so DELAY is always the latest the partial can arrive. At full amount both cover eight octaves, a range of 256 to 1. A partial set to a 2 ms attack rises in 2 ms when struck hardest and takes half a second when struck softest. A 400 ms delay stays 400 ms when struck softest and drops to under two milliseconds when struck hardest. The lengthened attack stops at five seconds, the longest the ATTACK row goes. STRIKE moves the whole row, including the part below the shortest time. A partial set to start at its peak starts there when struck hard, nearer the zero crossing when struck more lightly, and past it into a short attack when struck softest. A quarter turn is worth two octaves of attack, so at full amount the softest note turns a start at the peak into an attack of about 13 ms.

Because it is set per partial, it does more than a velocity curve for the whole instrument. Give the upper partials a strike amount and leave the fundamental at zero. A hard chord then arrives all at once and bright, while a soft one builds up, with the fundamental first and the upper partials after it.

One useful shape is the one a music box or a thumb piano has. Set no sustain at all, so the partial decays to silence while the key is still down, and then set a key-off level that brings it back. Decaying to zero does not end the note.

## Ganging the channels

**One button at the top of the caption column on the left** sets what a drag in the mixer does. Instead of a word, it shows the pointer you will get: a plain arrow, an arrow with two chain links, or an arrow with a drawn contour. Click it to see the three by name. A right-click anywhere in the mixer opens the same menu.

**Link** gangs the strips: dragging any knob moves the same knob on the other strips. It works relatively. It adds an offset to each strip's current value instead of setting them all to one value, so a spectrum you have shaped by hand keeps its shape. The offset is always measured from the values at the start of the drag, so moving the knob back to where you started restores the strips exactly, even if some of them hit the end of their range along the way.

Which channels a drag reaches, and how the movement is shared between them, are set in the same menu, under the three tools. With Link chosen, pointing at a knob lights every knob a drag from it would move, before you click. How brightly each one lights follows the curve.

**Draw** is the third tool. Dragging across the faders sets each channel you pass over to the pointer's height, so you can draw a spectrum in one stroke instead of one fader at a time. Holding shift does the same whatever tool is chosen, and the button shows the contour while the key is down. Either way, the button shows what a drag would do.

Drawing is different from Link. Link shares one relative move across the channels by a rule, and drawing sets absolute values freehand, so neither can do the other's job. Channels the pointer skips over between one moment and the next are filled in, so a fast drag leaves no gaps, and the whole stroke is a single undo step. Drawing works on the faders only, because a fader's value is its height and a pointer's height means nothing for a knob.

While drawing is armed, every fader lights in the same accent colour as the button, so you can see what the next drag would reach before you click, the same way Link previews its knobs. The pointer becomes a pencil. If you hold shift with another tool chosen, the button shows the pencil while the key is down and goes back to your tool when you let go.

| Scope | Reaches |
|---|---|
| All | every partial |
| Same interval | only the strips sharing the interval class of the one you grab, so you can move just the fifths, or just the octaves |
| Odd harmonics | 1, 3, 5 and so on, the half of the series that sounds hollow |
| Even harmonics | 2, 4, 6 and so on |
| Selected | the channels you picked by their numbers, whichever they are |

To move a set of channels no rule describes, pick them. Cmd-click a channel's number on a Mac, or Ctrl-click on Windows and Linux, to pick it or put it back, and drag across the numbers to do the same to every channel you cross. A drag that starts on a picked channel takes channels away instead. Picked channels show a short bar under their number, and picking one switches to Link with the Selected scope, so the next drag moves only those channels. The selection is kept with your session, and Clear selection in the Link menu empties it.

| Curve | How the drag is shared out |
|---|---|
| Uniform | every selected strip moves by the same amount |
| Taper from the grab | a strip moves less the further along the series it sits from the one you grabbed |
| Spread / gather | dragging up spreads the strips apart in random directions, dragging down pulls them together onto the strip you are dragging |

Taper is centred on the strip you grabbed, which always moves by the full amount of the drag, because the knob under the mouse has to follow the mouse. From there, each strip's share falls off in a straight line with distance, reaching nothing thirty-one channels away.

So the channel you grab sets the shape. Grab channel 1 and the movement tilts down across the full width of the mixer, from the full amount to nothing. Grab channel 32 and it tilts up instead. Grab channel 10 and the curve peaks at that channel and falls away on both sides, to 0.71 at channel 1 and 0.29 at channel 32. Distance is measured against the full width, not against the nearer end, so the reach is the same wherever you grab. Scaling it to the nearer end would make a grab on channel 2 drop channel 1 to nothing in a single step.

LINK gangs the 32 partials only. The noise channel has no tuning, pitch modulation or drift to gang.

## Working the panel

**Double-clicking a knob resets it to its default**, which is that control's own default, not the bottom of its range. On a fader that is 0 dB, not silence.

**Undo and redo** matter more here than on a smaller instrument, because one LINK drag moves the same knob on all 32 channels. Without undo, the only way back would be to reload the preset. Each step is a whole gesture, not a single value change, so that drag is undone in one step, not thirty-two. Cmd-Z and Cmd-Shift-Z, or Ctrl with the same keys on other systems, work when the host passes them on. Many hosts do not, because a DAW usually keeps those keys for its own undo. Both are also near the top of the Settings menu, and that always works. Loading a session starts with an empty undo history.

**Hovering over a strip** shows the partial it belongs to: the harmonic number, the interval that harmonic makes with the fundamental, how many semitones above the played note its tempered position is, and how far the exact ratio is from that in cents. It is that channel's row in [the table of the series](https://benjamindehli.github.io/overtonium/tuning/), shown in the mixer. The mute and solo buttons show which harmonic they would silence.

**Right-clicking anywhere in the mixer** opens the LINK menu. On the mute and solo buttons, a right-click offers to mute or solo only that channel, or to clear every mute or every solo at once, instead.

**Option-click or Alt-click** on M or S mutes or solos only that channel, and turns the switch off on every other channel, the noise channel included. Doing it again on the channel that is already alone turns it off, so the same click goes in and out of hearing one partial. Either way it is a single step to undo.

## Reading the mixer

### The two readouts

Each channel shows two figures: the cents set by its tuning knob and the level of its fader. Both are drawn as seven-segment displays instead of text, so a number the instrument reports looks different from a number you typed. Two of the readings are not numbers. A partial left in equal temperament reads **Et**, and a fader all the way down reads **-inF**. The fundamental and the octaves read **0.0**, because their just interval is the note itself. As on a real display, each has a fixed row of cells whose segments are either lit or unlit, so the figures are always the same size.

### Lamps on the rules

Each of the rules that divide a strip into groups has a lamp. It shows what the group below it is doing to this partial at that moment.

| Rule | Shows |
|---|---|
| PITCH MOD | a needle on a track, flat to the left and sharp to the right, showing where modulation, drift and any glide still in progress have put the partial |
| ENVELOPE | how far up its envelope the partial is, while the key is down |
| KEY OFF | the same, once the key is up and the key-off stage has taken over |
| AMP MOD | how far the tremolo has lowered the level, so it pulses at the tremolo's rate and swings further at greater depth |
| OUTPUT | one long lamp in three parts, one for each row below it: the left shows how much of the partial velocity has left, the middle how much the keyboard row has left on the key being played, and the right what pressure on the key is adding |

The lamps read the voice pool, not the knobs, so they show what a note is doing, not a setting, and nothing pulses when nothing is sounding. They follow the loudest voice on that partial, the same one the meter follows. The output rule has one long lamp in three parts, one for each row below it and in the same order. VELOCITY can only reduce the level, so its lamp shows what the row leaves: full when velocity has taken nothing from the partial, dark when it has taken all of it, and full at every velocity when the knob is at the centre. KEYBOARD reads the same way, for the key being played, and follows a glide across. AFTERTOUCH adds level instead, so its lamp is dark at rest and lights up as you press into the key, whichever direction that row is set to push. The needle's scale is fixed and the same on every strip, so you can compare two channels by eye. Full deflection is 225 cents. That covers a vibrato and a drift at full depth, but not the widest the depth knob reaches. A square set to jump an octave pushes the needle to the end of its track, and so does a wide glide until it is within a couple of semitones of its note. With a scale wide enough for an octave, an ordinary vibrato would not move the needle at all.

### Meters

Each channel meters the actual output of its partial, on a decibel scale down to -48 dB. The envelope, tremolo, velocity, aftertouch and mute or solo all show in it, and you can watch the spectrum change as a note decays. The meter fills the fader's own track instead of sitting beside it, and the fader cap is drawn as translucent glass so the meter shows through it. It shows the loudest instance of the partial across the sounding voices, not the sum, so it shows the shape of the patch instead of hitting the top as soon as you play a chord.

### Colour and hover

Channels are coloured by interval class, in chromatic order along a narrow range from blue at the octave, through magenta and red, to yellow at the major seventh, so you can pick out octaves, fifths and thirds while you scroll. The range is narrow on purpose, so 32 channels look like one family instead of a rainbow, and none of the colours overlap the green and cyan of the global controls. Pointing at a knob highlights its row across the whole mixer and marks its channel, so the control where the two meet is the one a drag would move.

### Silenced channels

A channel you cannot hear, because it is muted or because another channel is soloed, turns its lights off. Its knob rings, both readouts and the macro columns beside its fader go dark, and its waveform screens switch off. The pointer on each knob and the fader cap still show where they are set, and M and S stay lit, since one of them is usually what silenced it.

### Folding the mixer down

Clicking a section heading in the caption gutter folds that group of rows away on every channel at once. Clicking that section's rule on any strip does the same, so you do not have to move to the gutter and then find the partial you were working on again. Pointing at a rule shows the same hand pointer as a heading and lights that section's caption and its disclosure mark, so you can see what the click would fold before you click. The window shrinks by exactly the height of those rows. Pitch modulation, envelope, key off, amp mod and output can each be folded. The tuning at the top and the faders at the bottom always stay. A folded heading keeps its activity lamp, so a hidden group still shows whether it is doing anything. The fold state is saved with the session, not with the patch.

## The noise channel

A noise channel, marked with a noise trace where the others have a number, sits on the right, after the series. It has the same envelope, tremolo, velocity, keyboard, aftertouch, pan, mute, solo, level and meter as a partial, and takes part in solo alongside them. It is not part of the harmonic series, so it has no tuning, pitch modulation or drift of its own. COLOUR takes the place of the tuning row, and a string takes the rest.

COLOUR tilts the noise from a dark rumble at one end, through flat in the middle, to a bright hiss at the other. The middle is fully flat, not just less filtered. Each voice runs its own noise stream, so playing a chord does not layer eight copies of the identical signal.

**A string** sits under COLOUR, where a partial's pitch modulation would be, for the noise to strike. It is a Karplus-Strong string: a delay line one period of the note long, fed back through a gentle low-pass filter, so a burst of noise sent into it comes out as a plucked note that loses its high frequencies first. It is tuned to the note and follows the pitch wheel, glide and Adaptive like the fundamental does. The noise strikes it after COLOUR and the envelope have shaped the noise, so a darker colour strikes it more softly. STRING crossfades from the noise alone to the string alone. DECAY sets how long the string rings while the key is down, DAMP how quickly it loses its high frequencies, and POINT where along the string the noise strikes. POINT runs from near the end, which keeps every harmonic, to the middle, which removes the even ones.

The envelope sets how the string is played. A noise envelope a few milliseconds long plucks the string and lets it ring, like the start of a harpsichord note, or of a guitar note under the organ. A noise held on a short, dark string blows into it, which gives a pitched breath. When the key is released, the channel's RELEASE damps the string. At the top of the keyboard, where the string is too short to tune, it fades back to plain noise instead of playing out of tune. COLOUR and the string's four knobs belong only to the noise channel, so pointing at them does not light a row across the mixer. The [Space Cowboy](https://benjamindehli.github.io/overtonium/presets/#what-ships) preset plucks it for its tremolo guitar, and KarplusStrong strikes it beside a lone fundamental.

---

Page: Overtonium effects: character, echo, reverb, converter and CLIP
URL: https://benjamindehli.github.io/overtonium/effects/
Last changed: 2026-11-07

# Overtonium effects

Everything on the bar along the top of the window affects the whole instrument, not a single channel: the oscillator character every partial is built from, the wobble under the series, the resonator, the drive on the bus where the channels are summed, the echo, the reverb, the converter, and the output stage with its meter.

## The signal chain

[Diagram: The order the signal passes through. Each of the thirty-three channels runs a sine partial through its envelope and key-off stage, its tremolo, its velocity and pressure amounts, its fader and its pan. The thirty-three are summed and run through the converter, the resonator and the bus drive. The mix then passes through the wobble, the echo, the reverb, the master fader and the safety clipper before it leaves. each of the thirty-three channels sine partial n envelope and key off tremolo velocity and pressure fader pan then the mix of all of them converter rate, bits resonator 88 strings bus drive wobble echo reverb master fader clipper summed out.]

The order things happen in. Everything on a channel strip runs before the channels are mixed, and everything on the bar runs after. This is why a channel meter and the output meter disagree while a tail is ringing.

[Image: The top bar: four icon buttons in two rows, settings and macros above the buttons that open the Playing and Tuning bars, then the wordmark over the loaded preset on a fourteen-segment display, the oscillator character, the stretch, track and wobble controls, the resonator with its mode and three knobs, the echo, the reverb, and the output meter with the master fader across it, two seven-segment converter readouts reading kilohertz and bits, and the output stage reading Soft.]

## Character

Chooses the oscillator that every partial uses, and is the first control in the group that sets up the series. All twelve are sine oscillators. They differ in how each one fails to make a perfect sine. **Pure** is an ideal sine that no real circuit produces. **Bulb** is a Wien bridge oscillator kept steady by a lamp. It adds no harmonics at all, but its level lags half a second behind every change of pitch, so it reacts to gestures. A vibrato is heard in the level as well as in the pitch, and a bend dips and then recovers as the lamp catches up. A held note with nothing moving sounds exactly like Pure, because the circuit's job is to hold the level steady. **Rail** is a phase-shift oscillator driven into its own supply rail. This hardens the tone with a third harmonic and no even harmonics. **Diode** is a triangle shaped into a sine by two mismatched diodes. This puts a kink in one half of the wave and adds a second, third and fourth harmonic. **Valve** is a triode biased so that one half of the wave bends over before the other. It is the only one here whose loudest added harmonic is an octave above the partial rather than a twelfth, and that is what people mean by valve warmth.

**Op-amp** is an amplifier that cannot keep up with a fast note. It is the one that changes as you play. Below a kilohertz nothing is limited and the partials are untouched. Above it they lose first their corners and then everything but their slopes, which leaves a triangle. So the top of the series gets harder as it climbs while the bottom stays the same, and a chord played high sounds like a different instrument from the same chord played low.

**Wheel**, **FM**, **PD**, **6581**, **VCO** and **Combo** are measured from recordings of real instruments, not simulated. **Wheel** is the tone generator of a tonewheel organ. Every pitch has its own wheel, so Wheel has seven sets of harmonics and picks one by the pitch of each partial. The low wheels have a strong second harmonic, -13 dB on the lowest C, which gets weaker up to A3 and then drops by fifteen decibels by C4. Above that the wheels are close to pure sines. The wheels also move slightly as they turn, as on the organ. Each one changes in pitch by about a cent and in level by about a tenth of a decibel once per turn of its shaft, between 16 and 31 times a second depending on the note, and the motor that drives them all moves every wheel together by about 0.9 cent at 25 Hz. A held note also carries faint tones from the wheels near it in the organ's generator: the same note one and two octaves down, and the wheels a fourth and a fifth below, 48 to 55 dB under the note. They follow the note's level and stop at C1, the organ's lowest wheel. **FM** is one operator of a frequency modulation synthesizer with nothing modulating it, and **PD** is the oscillator of a phase distortion synthesizer with the distortion turned off. Both are nearly pure, with a second harmonic near -54 and -52 dB and weaker ones above it. That is about thirty decibels below Rail, so the difference from Pure is small, but it is there. Each note on them also carries the noise measured under the instrument's own notes, about 62 dB under the note on FM and 69 dB on PD, so they are not perfectly clean either. It follows the notes, so there is no hiss when nothing plays. On FM most of the hiss above 6.4 kHz is a fixed part that follows only the loudest note, as on the instrument, so it hardly builds up as you add keys: five keys add about 3.6 dB of hiss up there, where the notes themselves add 7.

**6581** is measured from a 6581 sound chip, which has no sine at all. It makes every waveform as a 12-bit number and turns it into a voltage with a ladder of resistors whose bits are not weighted exactly. Each bit is worth about 1.95 times the one below instead of twice, so where many bits change at once the output falls back instead of stepping up. That uneven staircase, and a slight bow after it, were measured from the chip's slow sawtooth and triangle, and the sine it would make is worked out from them. It has a second harmonic at -23 dB and a long tail of odd harmonics between -55 and -63 dB, which is the chip's grit, carried to the 32nd. Part of the bow comes after the chip's voices are mixed, so a chord on 6581 also gets faint sum and difference tones between its notes, as the chip does. It also has the chip's hiss, which stays the same however many keys you hold. The chip was recorded through [SidStation ASID](https://benjamindehli.github.io/sidstation-asid/), a plugin by the same maker that writes the chip's registers directly, so every setting reached the chip exactly as set and each voice could be recorded on its own.

**VCO** is an analogue voltage-controlled oscillator's sine, recorded straight from its output, and the mixer it feeds. The sine is made from a triangle, so it has a second and a third harmonic near -25 dB and weaker ones above them, the same at every pitch. The mixer was recorded driven from quiet to well past clipping. It stays clean until it reaches a rail and then stops, and its negative side reaches its rail about two and a half times sooner than its positive side. On the bus that is what VCO does when you play hard: a chord runs into the negative rail first, which adds an octave as well as a twelfth. Each note also carries the oscillator's own faint hiss, about 122 dB per hertz under the note, and each partial's pitch wanders by about a fifth of a cent, as the recorded oscillator's did, so a held chord is never perfectly still.

**Combo** is the 8' footage lever of a transistor combo organ. The organ divides one oscillator down for every octave of a note, which gives a square wave, and the lever is that square through a filter. So it has only odd harmonics, and the filter stays put while the notes move, so the low notes keep more of the square: the third harmonic is -16 dB on the lowest C and -35 dB above C7. It has eight sets of harmonics and picks one by the key you play, because the organ filters each key once for all its levers, so the higher partials of a note have the tone of the organ's higher levers on that key. Its notes mix cleanly, so it has no bus drive, and it carries the organ's faint hiss, about 98 dB per hertz under the loudest note, which does not build up as you add keys.

Because this is an additive instrument, these harmonics land among the partials you are already mixing, at exact multiples of the partial that made them. The rack's own spread, described below, moves the real partials off those exact ratios, and TUNE and STRETCH move them further. So the two beat against each other instead of adding up, and that is what makes a character sound like a circuit and not like a tone control. The harmonics are removed as a partial climbs, so nothing folds back down. A partial near the top of the range plays a plain sine whatever character is chosen.

Each character is also a rack of thirty-two oscillators, not one oscillator used thirty-two times. Units built to the same spec do not come out identical, so each of the five circuits puts each partial a couple of cents and a fraction of a decibel off its target. The tonewheels are all driven by one motor through gears, so Wheel keeps every partial in tune and only varies the level, by up to a decibel. FM, PD and 6581 have no spread at all, like Pure, and neither does Combo, because the organ makes every octave of a note from one oscillator. Only one VCO was recorded, so it takes the spread of the Op-amp circuit. The amount is fixed for each character, so you get the same rack every time you load the patch, in every host and on every machine. This is not DRIFT, which wanders. Partial 1 is the reference the other thirty-one were tuned against, and it stays exactly in place. So a note still lands on the key you played, and choosing a character does not change how loud the patch is. This spread is what makes the beating above happen in a patch nobody has detuned. The units also differ in how much they distort: in the five circuits, each unit is built to one of three drives, fifteen percent either side of the nominal.

The button lights in the character's own colour, so you can tell which one is chosen at a glance. FM, PD and 6581 light in a colour of the instrument they were measured from: FM in the green of the synthesizer's membrane buttons, PD in the colour of the model name printed on the synthesizer, and 6581 in the blue of the startup screen of the computer the chip was made for. The others are spread as far apart in hue as they go. Pure does not light, because it adds nothing.

Character is part of the patch, not part of the setup, so a preset can change it and saving a preset stores it. Thirty-nine of the forty-four [factory presets](https://benjamindehli.github.io/overtonium/presets/) use one, so most of what you load already uses a circuit or a recording, not the plain sine.

## Wobble

Makes the whole instrument sound like a warped record. The pitch is bent by reading the output back through a delay line whose length keeps changing, as happens when a platter runs off-centre or a capstan slips. There is a slow warp at about once per turn of a 33 rpm disc, a faster wander on top, and now and then a sharp slip that bends hard and settles back. The slips grow in both rate and size with the square of the control, so a low setting sounds like a tired turntable and a high one like a badly failing tape transport. It comes before the echo, so the repeats carry its wobble instead of wobbling on their own.

## Resonator

The resonator is eighty-eight sympathetic strings, one for each key of a piano from A0 to C8, that resonate with what you play. Each is a Karplus-Strong string, the same kind the noise channel strikes. They are fed the sum of the thirty-three channels, and what they ring back is mixed in before the bus drive. So the drive and everything after it treat them as part of the instrument, not as a room around it. It has its own frame on the top bar, after the character's. The RESONATOR button shows the mode and opens a menu of the five, and it is lit whenever the mode is not Off, in a colour for each mode. The ECHO, REVERB and CLIP buttons do the same, each machine in its own colour, so the bar shows what is chosen without reading it. The knobs beside each button light in the same colour.

**Piano** works like a piano: the strings of the keys you hold are free, the rest are damped, and the sustain pedal lifts every damper. The sostenuto pedal keeps the dampers up for the keys that were down when it went down. With the sustain pedal part way down, the dampers rest lightly on the strings, so they ring for a time between a damped string's and DECAY. A held key's string is tuned to the pitch the key is sounding. The other strings wait at the pitch a key struck now would sound, which under Adaptive is the tuning their pitch class had last. None of them bends, because a piano's strings do not follow the wheel. **Open** is Piano with the sustain pedal always down: every string rings, and held keys ring the loudest. **Octave** is twelve strings from C3, always free, so every note finds something to ring with. **Series** is the tuning root's first eight harmonics from its second octave, tuned pure and always free. It works as a drone that responds to any note that falls on the root's own harmonic series. The root is set on the [Tuning](https://benjamindehli.github.io/overtonium/tuning/#root) bar or in Settings, and it stays available for Series under equal temperament too. **Fifths** is a drone of the root and the pure fifth above it in four octaves, from C2 to G5 on a root of C, always free. It follows the root as Series does. **Chord** is Piano that keeps the last chord: the strings of keys struck together stay free after you let go, so you can play a chord, let go and play a line over it while the chord rings on. A key added while the chord is down joins it, a new chord struck after letting go replaces it, and a single note leaves it be. The strings are as stiff as [STRETCH](https://benjamindehli.github.io/overtonium/tuning/#stretch) makes the partials, so their overtones are spread the same way and a stretched partial still rings its string. Each string rings back on the side of the partials that drive it, so a panned patch keeps its resonator with it, and a patch in the middle rings in the middle.

**MIX** goes from the instrument alone to the resonator alone, as on the echo and the reverb. On a held chord the level stays about the same as you turn it up. At 100 % only the resonator is left, so short notes lose their attack there. **DECAY** sets how long a free string rings, from 200 ms to 10 s. The level stays about the same at any setting. **DAMP** sets how quickly the high end dies away. A damper stops its string over 0.8 s at the bottom of the keyboard and 0.25 s at the top, or sooner if DECAY is shorter, so a released bass key leaves the short hum of a felt damper on a heavy string. A string freed only by the pedal gets much less of the signal than a held key's string, so the pedal adds a bloom under a chord instead of making it louder.

The resonator is part of the patch. Metallic Piano, Space Cowboy, KarplusStrong, Tape Pluck, 8-bit Choir, Soft Trio, Pendulum and Chapel use it, and the other factory presets have it off. With eight voices it uses about one percent of a core with the pedal up, and about five percent with every string freed by the pedal.

## Bus drive

The thirty-three channels are not just added together. They are summed onto a bus with drive, by an amount and in a style set by the oscillator character. So a patch gets louder into the same kind of saturation a mixing desk would give it. It responds to level, not to a setting: play quietly and it is barely there, play hard and you hear it. Pure and FM sum cleanly and add nothing. PD's bus has the slight bow measured in the phase distortion synthesizer's own mix, which gives each note its second harmonic and gives a chord faint sum and difference tones between its notes, as the instrument does. The 6581's bus has the part of its chip's bow that was measured on the mix of its voices. There is no knob, because the amount is part of the character you choose.

## Echo

Three machines behind one button, and the button shows which one is running. **Tape** is a loop with two motors. Moving the time knob makes the tape speed up or slow down to the new setting instead of jumping to it, so the repeats slide in pitch on the way. The two paths run at slightly different speeds, so a repeat comes back doubled rather than bounced. **BBD** is a bucket brigade delay. It loses its high end because a bucket brigade is a sampler with almost no filtering, and it sweeps as the clock wanders. **Digital** is a clean delay: what goes in comes back out, later and quieter, crossing to the other side on every repeat. It is the only ping-pong of the three.

When the echo is off, its knobs keep their settings, but their rings go dark and only the pointer on each knob shows where it is set. The reverb and the resonator do the same when they are off.

AGE means something different on each machine. On the tape it is wear: the high end lost on every pass, how far the motor wanders, and how hard the tape saturates. On the BBD it is the clock slowing down and how much the companding breathes. On the digital it is bits and sample rate, so a tail starts as the clean signal and ends as a degraded copy of it. The tape never reaches zero wear, because a transport that held speed exactly would put the repeat back in mono, and no real transport holds speed exactly.

## Reverb

Three machines again, on the same kind of button. **Room** is a feedback delay network whose room size follows its decay time, because a long tail in a small room sounds like a spring, not a space. **Modulated plate** is a sheet of steel with no shape you can hear, so a hit spreads across all of it at once and the tail is dense from the start instead of building up in stages. Its allpass filters move further than a real plate's would, which is why the menu calls it modulated. Without that movement, a long decay repeats its own pattern instead of washing out. **Spring** is a tray of coiled springs. A coil carries high frequencies faster than low ones, so a hit comes back as a chirp that starts high and falls. Every bounce adds another chirp, so a spring sounds less like what you played as it decays, not just quieter.

DECAY, DAMP and PRE mean the same on all three and are matched, so switching machines changes the character of a tail but not its length or loudness. On the plate, DAMP is the sheet losing its high end to the air. On the spring, it is the spring's own loss each time the signal crosses the tray.

## The converter

Shown as two seven-segment readouts under the output meter. Turning the sample rate down really lowers the rate, it is not a filter on top. The whole voice pool renders at the lower rate, which costs a fifth as much at 8 kHz, and partials above the new ceiling fold back down instead of disappearing.

## CLIP

Eight machines behind one button, on the finished output after the master fader, so the master fader sets how hard they are driven. The fader and the choice are both saved with the preset, so how hard a patch drives this stage is part of the patch. **Soft** is linear up to seven tenths and then rounds off, gradually and symmetrically. It is the default, and about half the factory presets use it. **Hard** is a wall at full scale: nothing happens until the signal reaches it. It sounds like a converter with no headroom left. **Asymmetric** lets one half of the wave bend over before the other, as a single-ended stage does. Soft and Hard treat both halves the same and add odd harmonics. Asymmetric adds even ones, so it sounds different, not just louder. **Limiter** does not change the shape of the wave. It turns the level down as the signal approaches the ceiling and lets it back up afterwards. **Fold** folds the wave back on itself instead of flattening it. This makes far more harmonics than the others, and they move with the level, so it sounds tonal where the others crunch.

**Console** is a console's mix bus, measured from a real chain at the settings it is used at: a transformer, a bus compressor and a tape stage, into the Limiter with its ceiling at -0.1 dBFS. The transformer thickens the lows as you drive it, because its core saturates on low frequencies long before high ones. The compressor is 4:1 and linked, with its sidechain filtered at 120 Hz so the bass does not pump it, and an automatic release that lets go fast after a short hit and slowly after a long one. The tape stage adds a steady third harmonic once it is driven, the lows first. It changes over time as well as by level, because a compressor does, and it is held at the same level as the others for a quiet signal, so choosing it does not make a patch louder.

**Cassette** is a four-track cassette recorder at 9.5 cm/s with Dolby B on a normal tape, measured by recording a test signal onto it and playing it back, into the Limiter with its ceiling at -0.1 dBFS. The lows saturate first, so a bass line thickens before anything else does, and the top rolls off gently, -3.6 dB at 12 kHz. Dolby lowers the tape's hiss by about 10 dB, and on this machine it does not quite cancel itself, so quiet highs come back a few decibels duller while loud ones pass almost untouched. The speed wanders by about a tenth of a percent, with flutter at 68 Hz and a slow wow, so a held chord moves very slightly in pitch. The hiss is there while something plays, rising a little under loud highs as it does on tape, and gone when nothing does. It is held at the same level as the others for a quiet signal.

**Print** is Console printed to Cassette, as the outboard chain and the deck were patched: the transformer, the compressor and the tape stage, a limiter at -0.1 dBFS, then the cassette recorder, into the Limiter at the end. From moderate levels the console's tape stage adds its steady third harmonic, where the cassette alone is still clean. At the top the compressor turns the loudest parts down before they reach the deck, and the deck still saturates the lows first. The tape hiss sits 6 dB lower than on Cassette, as it does when the outboard chain puts the music onto the tape at a better level. It is held at the same level as the others for a quiet signal.

Off lets the output go past full scale and reach the host unchanged. Use it when something later in the chain catches the peaks. None of the eight lets the output past full scale. That limit is the purpose of the stage, not a side effect of the shapes, and each shape stays within it however hard it is driven. The button shortens Asymmetric to ASYM, Limiter to LIMIT, Console to CNSL, Cassette to CASS and Print to PRNT to fit, while the menu and your host's automation lane show the full names. It sits under the meter with the converter readouts, because together they are everything that happens to the signal on its way out. The switch is also in the Settings menu.

With **Limiter lookahead** switched on in the Settings menu, the Limiter looks two milliseconds ahead. The gain reduction is already in place when the peak arrives, instead of following it, and that is what makes it a true limiter and not a clipper. This adds latency. A plugin whose latency changed with the preset would make the host rebuild its processing graph every time, so the latency applies to all eight machines and to the stage when it is off. Overtonium reports 108 samples at 48 kHz, a little over two milliseconds, which your host compensates for on playback.

**It is off by default**, and the plugin then reports only the bus stage's 12 samples. Only the Limiter, Console, Cassette and Print use the lookahead and most factory presets use none of them, so with it on most patches would pay two milliseconds for nothing. The limiter still works with it off. The detector sees a single sample instead of a window, so the gain is exact and immediate and the ceiling still holds. But the smoothing is lost, and a gain that jumps in steps is a distortion in itself, about fifteen times as much. It is in Settings and not on the bar because nothing a preset can change is allowed to change what the plugin reports to your host. If you rely on the Limiter, Console, Cassette or Print and want them smooth, switch it on. Most hosts pick up the new figure while running, and some only when playback stops or on reload, so choose before a take, not during one.

## The output meter

Horizontal, and split into left and right. The two channels are identical until something is panned off centre, and a single summed meter would hide exactly that. It reads the finished output after master gain and the clipper. The lamps run from teal through amber to orange as they approach full scale. Each colour is fixed to its decibel position instead of stretching with the level. A scale below is marked at -48, -36, -24, -12, -6 and 0 dB. Each bar holds its peak for a couple of seconds.

---

Page: Playing Overtonium: MPE, automation, macros and MIDI Learn
URL: https://benjamindehli.github.io/overtonium/playing/
Last changed: 2026-11-07

# Playing Overtonium

How to play the instrument and control it from outside the window: MPE, legato and glide, the Playing and Tuning bars, the Settings menu, automation from a host with every parameter listed, macros that move a whole row from one control, and MIDI Learn for playing it from a controller.

## MPE

Off by default. It is in Settings and in the Playing bar. With MPE on, a controller that gives every note its own MIDI channel can bend and press each note separately. On this instrument, that means each finger gets its own copy of all 33 aftertouch destinations. Press into one note in a chord and only that note's upper partials come in.

Slide, the forward and back movement of a finger on the key, has its own destination setting in Settings and in the Playing bar. By default it changes that note's brightness, thinning or opening up the top of the series, which is what players expect from that axis. It can control the tuning instead, so that pushing one finger forward moves that note alone [between equal temperament and just intonation](https://benjamindehli.github.io/overtonium/tuning/) while the rest of the chord stays where it is. A keyboard without a slide axis sends nothing, so nothing changes.

Slide is read as movement, not as a position. The first value a note receives becomes that note's zero, and what affects the sound is how far the axis has moved from it since. This is because controllers handle the axis differently. A Seaboard reports where the finger is and rests at the centre. An Expressive E Osmose uses the first part of the key travel for pressure and only then starts sending slide, from the bottom of its range upward. Read as a position, the Osmose would jump the moment the axis starts sending. Read as movement, both start the note as the patch sets it and follow the finger from there.

The movement is scaled so that the longer of the two directions from the resting point covers the full slide, and both directions use that same rate. A resting point near the bottom of the axis leaves very little room below it. Giving that small range its own scale would make it far too sensitive: pressing in would feel smooth, but coming back out would jump straight to dark. So the short side has the same sensitivity as the long side and reaches less far.

MPE is a setting, not always on, because it changes what a channel number means. With it off, the channel is ignored: a key-up on channel 7 releases a note that started on channel 1, which is what a single-channel keyboard needs. With it on, the channel is part of the note's identity, so the same key can be held on two channels at once and bent in two directions, as two voices instead of one retriggering the other.

## Legato

A switch in Settings and in the Playing bar, beside the voice count. What it does depends on that count. **With one voice** the instrument is monophonic. Pressing a key while another is still held moves the note instead of starting it again, so the envelopes continue and a run keeps the shape the first key gave it, without restarting the attack. Releasing a key while others are still held goes back to the most recent of them, which makes trills work. Only releasing the last key releases the note.

**With more than one voice it plays a top line.** A key pressed above every held note except the highest one continues that highest note's phrase. It starts its own voice, which takes over the highest note's envelopes at their current point and glides from its pitch, so the sound does not start again. Every other key plays as normal. This means you can play a legato melody over a staccato chord. The chord retriggers each time its keys are released and pressed again, and the melody carries on from note to note above it, up or down, as long as it stays above the chord.

A legato melody over chords played below it, two notes to a bar, one with the chord and one in between. It is played softly on Synth Ensemble, where each new note takes about half a second to swell in. With Legato off, every note of the melody swells in again. With it on and eight voices, the melody carries on from note to note, with or without the chord, while the chords still swell in underneath.

**Legato off**

**Legato on**

The note a melody carries on from is never cut off, so a key that turns out to be part of the chord always sounds. It keeps sounding while its key is down and fades quickly when the key is released, instead of playing its full release under the melody. The same goes for the upper note of a trill when you return to the lower key.

A key pressed below the highest note could be the melody stepping down, or a chord played under a held melody. Only what the previous melody key does next would tell them apart, so the instrument goes by what is most likely. Above the highest note, a key always carries on. Below it, with a chord held and no other key just struck, it carries on if it is above that chord, however far down it goes. Alone under a single melody note, or among keys struck together, it carries on only if it is within a major third and is the highest of the keys struck with it. A melody usually moves by seconds and thirds, and the top of a chord usually sits further down. A wrong guess costs no more than a new attack. No key is ever delayed.

When the instrument runs out of voices, the melody is the last voice taken. The sustain pedal holds the sound but not the melody, so a note held only by the pedal does not carry anything on. With MPE on it works the same way, and a note carried onto a key on its own channel takes on that key's bend, pressure and slide.

Legato moves the note in a single step. With a [glide](https://benjamindehli.github.io/overtonium/playing/#glide) time on the channels, it slides there instead.

## Glide

**Every channel has its own glide**, the last knob under PITCH MOD. It sets how long that partial takes to move from the previous note to the new one, from Off, the default on every channel, to five seconds. Because each partial has its own time, the partials do not have to arrive together. Short times on the upper partials and a long one on the fundamental bring the brightness in first, with the pitch following. The other way round, the pitch arrives first and the tone follows. With settings in between, the partials stop agreeing on a fundamental while they move and settle on one as they arrive. A single portamento cannot do this.

The same phrase twice on the [Glide](https://benjamindehli.github.io/overtonium/presets/) preset: up a fifth, up a fourth and down the octave. With every glide at zero, each note arrives straight away. With the preset's settings, from nothing on the fundamental to two and a half seconds on the thirty-second partial, the pitch arrives at once and you can hear the upper partials still moving.

**Glide off**

**Glide on**

A partial glides in a straight line in pitch, and everything that depends on pitch follows the glide. The oscillator character is chosen for the frequency the partial has reached, and the Bulb character's lamp lags behind a long slide as a real lamp would. Vibrato, drift and both kinds of bend are added on top. The needle on each partial's PITCH MOD rule follows the glide and swings back to centre as the partial arrives, so with different glide times you can watch the partials arrive one at a time.

**Two switches control every channel at once.** They are the two letters **L** and **T** beside the GLIDE caption on the left of the mixer. They are also at the end of the menu that opens when you right-click any GLIDE knob, under *Glide (every channel)*. A letter is lit when its switch is on. The switches are part of the patch, so a preset saves them and a legato lead can set them. The letters show their state as soon as a preset loads.

| Switch | Choices | What it decides |
|---|---|---|
| When, L | Always, the default, or Legato (L lit) | Always glides from the last note played, even if its key was released long ago. Legato glides only when the new key is pressed while another key is still held, so a phrase played detached jumps to each note and the same phrase played legato slides between them. A note held only by the sustain pedal does not count as held |
| What the time means, T | Fixed rate, the default, or Fixed time (T lit) | Fixed rate makes the time the length of a slide over one octave, so a fifth takes a little over half of it and two octaves take twice as long, like a finger sliding along a string at a steady speed. Fixed time makes every slide take the set time however far it goes, so a leap and a step arrive at the same moment |

Played polyphonically, each new note glides from the last note played. In [Legato](https://benjamindehli.github.io/overtonium/playing/#legato), the voice carrying the phrase glides from wherever it is: the single voice when one voice is set, and the top line when more are set. A key pressed halfway through a slide reverses it from the current point instead of jumping to either end. A trill that returns to a key still held also glides back to it.

GLIDE is a row like any other, so [LINK](https://benjamindehli.github.io/overtonium/controls/#link) sets it across the series in one drag and a [macro](https://benjamindehli.github.io/overtonium/playing/#macros) can change every channel's glide time from one control.

## The Playing and Tuning bars

The lower two of the four buttons at the left end of the top bar, a keyboard and a tuning fork, each open a bar between the top bar and the mixer. The bars hold the settings you change while playing, as opposed to the ones you set once. Both bars start closed, and each instance remembers which are open. A bar takes its height from the mixer, which scrolls, so the window stays the same size. If the window is already at its minimum height, it grows by the height of the bar.

**Playing** has the voice count with IN USE, which shows how many voices are sounding, release tails included. Beside them are Legato and No stack, the pitch bend range, and MPE with Aftertouch from and Slide to. After these it shows what your hands are doing. SUSTAIN lights while the sustain pedal is down, with a bar along its foot for how far, and SOSTENUTO under it while the sostenuto pedal is. PITCH shows the pitch wheel's position from the centre against the bend range, or with MPE on, the key bent furthest. WHEEL and PRESSURE fill from the left. PRESSURE shows the keyboard's own pressure, from the channel or from the key pressed hardest, so the wheel moves only WHEEL even when Aftertouch from follows it. Last are the eight [macros](https://benjamindehli.github.io/overtonium/playing/#macros) as knobs in their colours. Each is named after the row it controls and turns by the same amount as its slider in MACROS. A macro not yet made shows a blank knob with its number. Right-click a knob to assign its macro, to edit it in MACROS, or to bind a controller to it. **Tuning** has the temperament, root and reference pitch, and under Adaptive, the drift beside them. Under Adaptive it also has a NO DRIFT switch that keeps the keyboard home, a 7:4 switch for the harmonic seventh, a PURE lamp and the Retune knob. Together they show [what Adaptive is doing](https://benjamindehli.github.io/overtonium/tuning/#adaptive) to the chord you hold. Last are twelve displays, one for each note, showing how far it is from equal temperament in cents, with the white keys on a lighter background as on a keyboard. Each setting's display opens the same list as its entry in Settings, and each switch lights while it is on, so a bar is another way to reach the same settings, not a separate copy of them. If the window is too narrow for a whole bar, the bar wraps onto more rows like the top bar does, and takes the extra height from the mixer.

The **sostenuto pedal**, controller 66 and the middle pedal on a grand piano, holds only the keys that are down when it goes down. Hold a chord, press the pedal and let go, and the chord keeps sounding while the notes you play over it stop as your fingers lift. Keys pressed after the pedal went down are not held, and a note held only by the sustain pedal is not caught. With both pedals down, lifting the sustain pedal leaves the caught notes sounding. With the resonator in Piano, the caught keys' strings stay free too. With MPE on it is read on the zone's master channel, like the sustain pedal.

The **sustain pedal** can be half-pedalled. A continuous pedal sends how far down it is, and the instrument uses all of it. All the way down it holds notes as it always has. Part way down, a note it holds fades once its key comes up, faster the less the pedal is down, and at the top of its travel the note fades as quickly as its own release. Moving the pedal changes the fade from that moment, so you can keep a line connected without blurring it. An on/off pedal, which sends only up and down, works as it always did. With the resonator in Piano, the dampers rest on the strings as far as the pedal is up, so the strings ring for less time too.

[Image: The Playing bar with C7 held on the pedal: eight voices with four in use, Legato off and No stack lit, a bend range of 2 semitones, MPE off with aftertouch from Either and slide to Brightness, SUSTAIN lit with its bar full and SOSTENUTO dark under it, the pitch wheel at rest in the middle, the mod wheel and pressure part way up, and three macros made, Level in red, Pitch mod depth in green and Release in blue, followed by five blank knobs numbered 4 to 8.] [Image: The Tuning bar under Adaptive with C7 held: the temperament reading Adaptive on a root of C at 440 Hz with a drift of 0, NO DRIFT off, 7:4 lit, PURE lit and the Retune knob near its start, then twelve displays for the notes, C, E, G and B flat lit at 15.4, 1.7, 17.1 and minus 15.8 cents, the white keys on a light ground.]

## The Settings menu

Settings you set once and then leave. The menu is where you set up the instrument, and the panel is where you make the sound, so where a control sits tells you which kind it is. The ones you change while playing are also in [the Playing and Tuning bars](https://benjamindehli.github.io/overtonium/playing/#bars).

| Entry | Choices | What it decides |
|---|---|---|
| Overtonium X is available... |  | Only there when a newer version is out, above the version. It opens the release on GitHub, and SETTINGS stays lit until you update |
| Which version this is |  | A heading near the top of the menu, not an entry, because it describes the build and is not a choice. A bug report asks for this number |
| Check for new versions | off by default | The only entry that connects to anything outside your computer. [What it sends](https://benjamindehli.github.io/overtonium/install/) is only a request for a file |
| Undo, Redo |  | The same steps as Cmd-Z, for hosts that keep that shortcut for themselves |
| Polyphony | 1, 2, 4, 6, 8, 12 or 16 voices. 8 by default | How many notes can sound at once |
| No stack | on by default | Playing a key again takes over the note it is already sounding instead of adding a second copy, as a string or a bar does. When off, every strike gets its own voice and the tails add up, so a bell struck repeatedly builds up |
| Legato | off by default | With one voice, a key pressed while another is held moves the note without starting it again. With more voices, the highest held note does this for a key above the rest, and everything below it plays as normal. See [Legato](https://benjamindehli.github.io/overtonium/playing/#legato) |
| Pitch bend range | 0, 1, 2, 3, 4, 5, 7, 12 or 24 semitones. 2 by default | How far the wheel bends. Only the commonly used ranges are offered, not all twenty-five |
| MPE | off by default | Whether a note's channel is part of its identity, which lets pitch bend and pressure work per note |
| Aftertouch from | Channel pressure, Mod wheel, Either. Either by default | What drives the AT row on every strip. Polyphonic aftertouch is not on the list because it is always routed, whatever this is set to |
| Slide to | Off, Brightness, Tuning. Brightness by default | What the forward and back movement of an MPE finger controls |
| Temperament, Root, Reference pitch | Nine temperaments, twelve roots, 415 to 466 Hz | What the keyboard is tuned to, as explained in [the tuning system](https://benjamindehli.github.io/overtonium/tuning/). Root is greyed out for equal temperament, which has no centre |
| Retune | Fast, 30 ms, the default, or Smooth, 300 ms | How long a held note takes to slide to its new tuning when the [Adaptive](https://benjamindehli.github.io/overtonium/tuning/#adaptive) temperament retunes the chord around it. These are two quick settings for the Retune knob in the Tuning bar, which runs from an instant jump at zero up to one second. Greyed out for every other temperament |
| Harmonic sevenths | on by default | Whether Adaptive tunes a minor seventh 7:4 above its root, the harmonic seventh, when nothing else in the chord would beat against it. When off, every seventh has the size the chain of fifths gives it. Also the 7:4 switch in the Tuning bar. Greyed out for every other temperament |
| No drift | off by default | Whether Adaptive keeps the chord you hold as close to home as it can. When on, a progression that would shift the keyboard by a comma moves a held note by a comma instead, and a key played after silence goes home. When off, the keyboard follows the music, up to two commas either way. Also the NO DRIFT switch in the Tuning bar. Greyed out for every other temperament |
| Phase reset | on by default | When a note begins from silence, starts each partial at the point its ATTACK sets: a rising zero crossing, or nearer its peak when the attack is turned below its shortest time. When off, a partial carries on from wherever its cycle was |
| Safety clip | on by default | Whether the output stage at the end of the chain is on. Best left on when you push 32 faders up. Which of its eight machines it uses is chosen on the bar, under the meter |
| Limiter lookahead | off by default | Whether the Limiter can look ahead at the incoming signal. On makes it smoother but adds two milliseconds of latency on every shape. Only the Limiter, Console, Cassette and Print use the lookahead, but the latency must not change with the patch. The only setting here that changes what the plugin reports to your host |
| Zoom | 75, 100, 125, 150, 175 or 200% | How big the window is on your screen. It sits at the bottom with the entry below it, because everything above these two is about the instrument and these are not |
| Fit all 32 channels |  | Resets the window to the width that shows the whole mixer, and to the height your unfolded rows need. A window remembers the size you dragged it to, so a narrow window opens narrow in every later session. This entry resets it |

Almost none of these settings are saved in a [preset](https://benjamindehli.github.io/overtonium/presets/). This is a rule, not a list that might get out of date: how you play the instrument and what it is tuned to belong to the session, not to any one sound in it. For the same reason, presets do not save the zoom, the window size or which groups of rows you have folded away.

The safety clip is the exception in this table, and the master fader is the exception outside it. The output stage has five shapes to choose from, and the master fader sets how hard the signal drives into it, so how hard a patch drives that stage is part of the patch, not of your setup. As a result, **loading a preset changes your output level**.

## Automating it from the host

Eight hundred and eighty-four parameters reach the host: forty-five global ones, twenty-four on each of the 32 partials, twenty-three on the noise channel, and six for each of the eight macros. They are listed in the same order as the panel, with the globals first, then each channel in turn, and the noise channel last.

Every per-channel name includes its channel, so you can search a parameter list either way. Type `H7` to find that partial, or type `Tune` to find that row across the whole mixer. A screen reader reads the same names, which is why they are words, not numbers.

| What you are after | What the host calls it |
|---|---|
| The master fader | `Master` |
| The seventh partial's tuning | `H7 Tune` |
| The eleventh partial's fader | `H11 Level` |
| The noise channel's brightness | `Noise Colour` |
| How long the echo's repeats are | `Echo Time` |

The noise channel has twenty-three instead of twenty-four because it has no tuning, pitch modulation, drift or glide, and has COLOUR and the four knobs of its string instead.

Everything in Settings is in the list too, even though no preset changes it. The temperament, the polyphony and the bend range can all be automated, so a host can change the temperament partway through a piece.

**Each parameter is identified by its own id, not by its position**, in the Audio Unit and in the VST3. This lets a release add controls without breaking automation you have already written. 1.6.0 added 65 new parameters among the existing ones, not after them, and every automation lane written before it still points at the same control.

LINK is not a parameter, because it is a mouse tool, not part of the sound. A LINK drag starts a gesture on every channel it reaches, writes each one, and ends them all together, so the host records what actually moved: thirty-two lanes if the drag moved thirty-two channels.

### Every parameter

All 884 as the host lists them, with a control that every partial or every macro carries given one row for all of them. The range and the default are what the plugin itself shows. The same list, one parameter per line with its id, is [parameters.json](https://benjamindehli.github.io/overtonium/parameters.json), written by the plugin rather than by hand.

**Global**, forty-five of them.

| What the host calls it | Id | Range | Default |
|---|---|---|---|
| `Master` | `masterGain` | -60.0 to 12.0 dB | -12.0 dB |
| `Polyphony` | `polyphony` | Legato, 1 voice, 2 voices, 4 voices, 6 voices, 8 voices, 12 voices, 16 voices | 8 voices |
| `No Stack` | `oneVoicePerKey` | Off or on | On |
| `Pitch Bend Range` | `bendRange` | 0 to 24 | 2 |
| `Phase Reset` | `phaseReset` | Off or on | On |
| `Safety Clip` | `safetyClip` | Off or on | On |
| `Lookahead` | `lookahead` | Off or on | Off |
| `MPE` | `mpe` | Off or on | Off |
| `Stretch` | `stretch` | -1200 ct to +1200 ct | Harmonic |
| `Tracking` | `track` | Off to 12.0 dB/oct | Off |
| `Character` | `character` | Pure, Bulb, Rail, Diode, Valve, Op-amp, Wheel, FM, PD, 6581, VCO, Combo | Pure |
| `Wobble` | `wobble` | 0 % to 100 % | 0 % |
| `Temperament` | `temperament` | Equal, Just (major), Pythagorean, Quarter-comma meantone, Werckmeister III, Young, Adaptive, Kirnberger III, Vallotti | Equal |
| `Tuning Root` | `tuningRoot` | C, C#, D, D#, E, F, F#, G, G#, A, A#, B | C |
| `Reference Pitch` | `referenceHz` | A = 415 Hz, A = 430 Hz, A = 432 Hz, A = 435 Hz, A = 438 Hz, A = 440 Hz, A = 442 Hz, A = 443 Hz, A = 444 Hz, A = 446 Hz, A = 466 Hz | A = 440 Hz |
| `Aftertouch From` | `atSource` | Channel pressure, Mod wheel, Either | Either |
| `Slide To` | `slideDest` | Off, Brightness, Tuning | Brightness |
| `Sample Rate` | `lofiRate` | Host, 32 kHz, 22.050 kHz, 16 kHz, 11.025 kHz, 8 kHz, 6 kHz, 4 kHz | Host |
| `Bit Depth` | `lofiBits` | Host, 16 bit, 12 bit, 10 bit, 8 bit, 6 bit, 4 bit, 3 bit, 2 bit | Host |
| `Echo` | `echoOn` | Off or on | Off |
| `Echo Mix` | `echoMix` | 0 % to 100 % | 25 % |
| `Echo Time` | `echoTime` | 20.0 ms to 2.00 s | 350 ms |
| `Echo Feedback` | `echoFeedback` | 0 % to 95 % | 35 % |
| `Echo Age` | `echoAge` | 0 % to 100 % | 29 % |
| `Reverb` | `reverbOn` | Off or on | Off |
| `Reverb Mix` | `reverbMix` | 0 % to 100 % | 25 % |
| `Reverb Decay` | `reverbDecay` | 200 ms to 20.00 s | 2.00 s |
| `Reverb Damping` | `reverbDamp` | 0 % to 100 % | 50 % |
| `Reverb Pre-delay` | `reverbPreDelay` | 0.0 ms to 250 ms | 0.0 ms |
| `Pitch Mod Shared` | `pmInPhase` | Off or on | Off |
| `Amp Mod Shared` | `amInPhase` | Off or on | Off |
| `Echo Type` | `echoType` | Tape, Bucket brigade, Digital | Tape |
| `Clip Type` | `clipType` | Soft, Hard, Asymmetric, Limiter, Fold, Console, Cassette, Print | Soft |
| `Reverb Type` | `reverbType` | Room, Modulated Plate, Spring | Room |
| `Glide Trigger` | `glideTrigger` | Always, Legato | Always |
| `Glide Mode` | `glideMode` | Fixed rate, Fixed time | Fixed rate |
| `Legato` | `legato` | Off or on | Off |
| `Adaptive Retune` | `adaptiveRetune` | 0.0 ms to 1.00 s | 30.0 ms |
| `Adaptive Harmonic Sevenths` | `adaptiveSeventh` | Off or on | On |
| `Adaptive No Drift` | `adaptiveHome` | Off or on | Off |
| `Resonator Mode` | `resonatorMode` | Off, Piano, Open, Octave, Series, Fifths, Chord | Off |
| `Resonator Mix` | `resonatorMix` | 0 % to 100 % | 50 % |
| `Resonator Decay` | `resonatorDecay` | 200 ms to 10.00 s | 4.00 s |
| `Resonator Damp` | `resonatorDamp` | 0 % to 100 % | 40 % |
| `Velocity Curve` | `velocityCurve` | Classic, dB | dB |

**On every partial**, twenty-four controls, each one on all 32 partials.

| What the host calls it | Id | Range | Default |
|---|---|---|---|
| `H1 Tune to H32 Tune` | `h01_tune to h32_tune` | Equal temp. to Just | Just |
| `H1 Pitch Mod Rate to H32 Pitch Mod Rate` | `h01_pmRate to h32_pmRate` | 0.01 Hz to 30.0 Hz | 4.00 Hz |
| `H1 Pitch Mod Depth to H32 Pitch Mod Depth` | `h01_pmDepth to h32_pmDepth` | 0 ct to 1200 ct | 0 ct |
| `H1 Pitch Mod Shape to H32 Pitch Mod Shape` | `h01_pmShape to h32_pmShape` | Sine, Triangle, Sawtooth, Reverse Sawtooth, Square, Square (up), Sample & Hold, Random | Sine |
| `H1 Glide to H32 Glide` | `h01_glide to h32_glide` | Off to 5.00 s | Off |
| `H1 Drift to H32 Drift` | `h01_drift to h32_drift` | 0 ct to 25 ct | 0 ct |
| `H1 Strike Velocity to H32 Strike Velocity` | `h01_strike to h32_strike` | -100 % to +100 % | 0 % |
| `H1 Delay to H32 Delay` | `h01_delay to h32_delay` | 0.0 ms to 5.00 s | 0.0 ms |
| `H1 Attack to H32 Attack` | `h01_attack to h32_attack` | 90° to 5.00 s | 5.0 ms |
| `H1 Decay to H32 Decay` | `h01_decay to h32_decay` | 1.0 ms to 20.00 s | 600 ms |
| `H1 Sustain to H32 Sustain` | `h01_sustain to h32_sustain` | 0 % to 100 % | 100 % |
| `H1 Key-off Swell to H32 Key-off Swell` | `h01_swell to h32_swell` | 0.0 ms to 5.00 s | 5.0 ms |
| `H1 Key-off Level to H32 Key-off Level` | `h01_offLevel to h32_offLevel` | 0 % to 100 % | 0 % |
| `H1 Release to H32 Release` | `h01_release to h32_release` | 1.0 ms to 20.00 s | 400 ms |
| `H1 Amp Mod Rate to H32 Amp Mod Rate` | `h01_amRate to h32_amRate` | 0.01 Hz to 30.0 Hz | 4.00 Hz |
| `H1 Amp Mod Depth to H32 Amp Mod Depth` | `h01_amDepth to h32_amDepth` | 0 % to 100 % | 0 % |
| `H1 Amp Mod Shape to H32 Amp Mod Shape` | `h01_amShape to h32_amShape` | Sine, Triangle, Sawtooth, Reverse Sawtooth, Square, Sample & Hold, Random | Sine |
| `H1 Velocity to H32 Velocity` | `h01_vel to h32_vel` | -100 % to +100 % | 0 % |
| `H1 Aftertouch to H32 Aftertouch` | `h01_aftertouch to h32_aftertouch` | -100 % to +100 % | 0 % |
| `H1 Mute to H32 Mute` | `h01_mute to h32_mute` | Off or on | Off |
| `H1 Solo to H32 Solo` | `h01_solo to h32_solo` | Off or on | Off |
| `H1 Pan to H32 Pan` | `h01_pan to h32_pan` | L100 to R100 | Centre |
| `H1 Level to H32 Level` | `h01_volume to h32_volume` | -inf dB to 0.0 dB | Differs by channel, 0.0 dB on H1 to -inf dB on H32 |
| `H1 Keyboard to H32 Keyboard` | `h01_keyboard to h32_keyboard` | 24 dB/oct above C-1 to 24 dB/oct below C9 | 0 dB/oct |

**On the noise channel**, twenty-three of them.

| What the host calls it | Id | Range | Default |
|---|---|---|---|
| `Noise Colour` | `noise_colour` | 0 % to 100 % | 50 % |
| `Noise Strike Velocity` | `noise_strike` | -100 % to +100 % | 0 % |
| `Noise Delay` | `noise_delay` | 0.0 ms to 5.00 s | 0.0 ms |
| `Noise Attack` | `noise_attack` | 0.2 ms to 5.00 s | 5.0 ms |
| `Noise Decay` | `noise_decay` | 1.0 ms to 20.00 s | 600 ms |
| `Noise Sustain` | `noise_sustain` | 0 % to 100 % | 100 % |
| `Noise Key-off Swell` | `noise_swell` | 0.0 ms to 5.00 s | 5.0 ms |
| `Noise Key-off Level` | `noise_offLevel` | 0 % to 100 % | 0 % |
| `Noise Release` | `noise_release` | 1.0 ms to 20.00 s | 400 ms |
| `Noise Amp Mod Rate` | `noise_amRate` | 0.01 Hz to 30.0 Hz | 4.00 Hz |
| `Noise Amp Mod Depth` | `noise_amDepth` | 0 % to 100 % | 0 % |
| `Noise Amp Mod Shape` | `noise_amShape` | Sine, Triangle, Sawtooth, Reverse Sawtooth, Square, Sample & Hold, Random | Sine |
| `Noise Velocity` | `noise_vel` | -100 % to +100 % | 0 % |
| `Noise Aftertouch` | `noise_aftertouch` | -100 % to +100 % | 0 % |
| `Noise Mute` | `noise_mute` | Off or on | Off |
| `Noise Solo` | `noise_solo` | Off or on | Off |
| `Noise Pan` | `noise_pan` | L100 to R100 | Centre |
| `Noise Level` | `noise_volume` | -inf dB to 0.0 dB | -inf dB |
| `Noise String` | `noise_string` | 0 % to 100 % | 0 % |
| `Noise String Decay` | `noise_stringDecay` | 50.0 ms to 10.00 s | 1.50 s |
| `Noise String Damp` | `noise_stringDamp` | 0 % to 100 % | 40 % |
| `Noise String Point` | `noise_stringPoint` | 0 % to 100 % | 25 % |
| `Noise Keyboard` | `noise_keyboard` | 24 dB/oct above C-1 to 24 dB/oct below C9 | 0 dB/oct |

**On every macro**, six controls, each one on all eight macros.

| What the host calls it | Id | Range | Default |
|---|---|---|---|
| `Macro 1 to Macro 8` | `macro1_amount to macro8_amount` | -100.0 % to +100.0 % | 0.0 % |
| `Macro 1 Row to Macro 8 Row` | `macro1_row to macro8_row` | None, Tune, Glide, Pitch mod rate, Pitch mod depth, Drift, Strike, Delay, Attack, Decay, Sustain, Key-off swell, Key-off level, Release, Amp mod rate, Amp mod depth, Velocity, Aftertouch, Pan, Level, Keyboard | None |
| `Macro 1 Scope to Macro 8 Scope` | `macro1_scope to macro8_scope` | All, Odd, Even, Prime/octave, Minor second, Major second, Minor third, Major third, Fourth, Tritone, Fifth, Minor sixth, Major sixth, Minor seventh, Major seventh, Selected | All |
| `Macro 1 Curve to Macro 8 Curve` | `macro1_curve to macro8_curve` | Uniform, Taper | Uniform |
| `Macro 1 Anchor to Macro 8 Anchor` | `macro1_anchor to macro8_anchor` | 1 to 32 | 1 |
| `Macro 1 Colour to Macro 8 Colour` | `macro1_colour to macro8_colour` | None, Red, Orange, Yellow, Green, Cyan, Blue, Violet, Magenta | Differs by macro, Red on Macro 1 to Magenta on Macro 8 |

## Macros

**A macro is one parameter that moves a whole row.** Macros exist for automation. A Link drag across the tuning knobs moves 32 parameters, and a host records only the last one you touched, so a change across the row that you can make by hand cannot be recorded. A macro gives that change a single control that a host can automate.

The icon of three faders on the top bar, beside the settings cog, opens the macros over the mixer. Each macro is one line: a colour, the row it controls, which channels of that row it reaches, how the movement is shared between them, and the amount. Add a macro uses one of the eight, and the cross at the end of the line frees it again.

**The eight macros are a pool, not a limit you are meant to notice.** A plugin declares its parameters once when it is created, and neither VST3 nor AU can add one later. So making a macro uses one of eight that already exist, and removing it frees it again. The panel shows the ones you have made, and your host's parameter list shows all eight whether you use them or not.

**Scope** is All, Odd, Even, any one interval, or Selected channels. Choosing fifths reaches harmonics 3, 6, 12 and 24. Selected channels reaches the channels you picked by their numbers with Cmd-click, or Ctrl-click on Windows, before opening the panel, and the menu lists them. The macro keeps its own copy, so picking other channels afterwards does not move it. To change them, pick the new ones and choose Selected channels again. They are saved with the session and with a preset of your own. **Curve** is Uniform or Taper. A taper has the most effect on a channel you pick, and less with distance on either side of it. Link's third curve, Spread, is not available and cannot be. It moves each strip in a direction based on your drag relative to the strip under the mouse, and a macro has no drag direction and no strip under the mouse.

**A macro offsets the patch's values instead of writing them.** Nothing in your sound is overwritten, the host has one lane to draw instead of thirty-two to record, and the knobs do not move when a macro does. That last part is normal for controls that work this way, but it can be surprising at first, so the controls show both values at once. The ring around a knob lights up to the value the macro has set, in the macro's colour, while the pointer on the cap stays at the patch's value, in the channel's colour. The gap between them is the macro. A fader has no ring, so it shows this in the gutters on either side of its track. Two narrow columns of segments appear there, in the same style as the meter behind the fader, and light from the bottom of the track up to the level the macro has set. The glass of the cap is tinted in the same colour.

With a macro at rest the two are in the same place, so a macro you have assigned but not moved looks the same as no macro.

On a [silenced channel](https://benjamindehli.github.io/overtonium/controls/#silenced) the macro's ring and columns go dark with the rest of the strip, and only the pointer and the fader cap still show the patch's value.

The amount runs from -1 to +1 whatever it controls, so a macro can always push a control either way from the patch's value. It moves each control by a proportion of its own range, not by a number of units. This matters most on a row like decay, which runs from a thousandth of a second to twenty seconds. Adding the same number of seconds would be a twentyfold change at one end and almost no change at the other.

A full sweep reaches the ends of the range, so if the patch value is already near one end, part of the macro's travel does nothing in that direction. This is on purpose. One macro reaches up to 32 controls that may be set anywhere, so no smaller part of the range would suit all of them.

**Right-click a macro's amount to bind a controller to it**, using the same MIDI Learn as the mixer's controls. Only the amount can be bound, because the row, the scope and the curve define the macro rather than move it. Its knob in the Playing bar has the same option, with Edit macro beside it to open MACROS. For a macro not yet made, the knob offers Assign macro instead, which creates the macro and opens MACROS. The macros icon lights whenever any macros exist, because otherwise you cannot see a macro's effect without looking at a ring.

## Playing it from a controller

**Right-click any control and choose MIDI Learn**, then move something on your controller. A violet ring marks the control while it waits, so you can see which control is listening without opening the menu again. Only one control can wait at a time. Right-click the same control afterwards and the menu shows which controller moves it, with options to remove the binding or bind a different one. In the mixer these items are below the LINK settings in the right-click menu. On the top bar, the Playing and Tuning bars and the noise channel, they have a menu of their own, because none of these is part of the series LINK gangs.

Your host almost certainly offers the same thing. The difference is that here you set up the mapping by right-clicking the knob itself, instead of finding `H17 AM Rate` in a list of 884 parameters, and it works the same way in every host.

**Eighteen controllers cannot be bound**, because the instrument already uses them and losing them would affect your playing. Controller 1 is the mod wheel, which can drive aftertouch. 64 and 66 are the sustain and sostenuto pedals. 74 is the slide axis, the easiest to bind by accident because it moves whenever a finger moves on an MPE keyboard. Controllers 6, 38, 98, 99, 100 and 101 are how a keyboard tells the instrument what kind of keyboard it is. 120 to 127 are the channel mode messages, which stop notes, reset controllers and switch modes. Some hosts send 121 when the transport stops. If you arm a knob and then touch the slide to steady a note, the note is steadied and the knob keeps waiting.

**It works the same with MPE on**, which takes extra handling. With MPE on, nearly every message goes to the part of the instrument that reads your keyboard's layout, so bindings are checked before that, not after. Otherwise MIDI Learn would not work on the controllers the instrument is most often played from.

The mapping is saved with your project, along with the window size and the folded sections, and **a preset never changes it**. Which controller moves which knob depends on your hardware setup, not on the sound, so a patch should not change it. A binding stores the parameter by name, so a later version that adds controls cannot quietly point it at something else.

Two things to expect. A control jumps to the controller's current position instead of waiting for the controller to reach it, so a hardware knob turned to the top takes the parameter there the moment you touch it. And undo does not reverse changes made from a controller, for the same reason it does not reverse automation.

---

Page: Overtonium presets: the forty-four that ship, and saving your own
URL: https://benjamindehli.github.io/overtonium/presets/
Last changed: 2026-11-08

# Overtonium presets

Forty-four presets ship with the plugin. All but a handful were set up on the panel and converted directly from the saved file. What ships is exactly what was played, not an approximation written afterwards.

## Hear them played

Eleven of them, played live, not rendered. The clips further down are one chord each, the same chord every time. That makes them easy to compare, but it does not show what a preset does when someone plays it. The video shows that.

Played on an Expressive E Osmose, with nothing on the output but the plugin's own effects.

[Video: Overtonium sound examples](https://www.youtube.com/watch?v=L1oYdPxGlGA)

Straight to one of them:

- [0:00 Synth Ensemble](https://www.youtube.com/watch?v=L1oYdPxGlGA&t=0s)
- [2:04 Metallic Piano](https://www.youtube.com/watch?v=L1oYdPxGlGA&t=124s)
- [2:55 Dream Phase](https://www.youtube.com/watch?v=L1oYdPxGlGA&t=175s)
- [4:04 FM Piano](https://www.youtube.com/watch?v=L1oYdPxGlGA&t=244s)
- [5:47 Sparkle Pad](https://www.youtube.com/watch?v=L1oYdPxGlGA&t=347s)
- [6:56 Dire Dire EP](https://www.youtube.com/watch?v=L1oYdPxGlGA&t=416s)
- [7:06 Music Box](https://www.youtube.com/watch?v=L1oYdPxGlGA&t=426s)
- [7:25 Stepped](https://www.youtube.com/watch?v=L1oYdPxGlGA&t=445s)
- [8:32 Wurli](https://www.youtube.com/watch?v=L1oYdPxGlGA&t=512s)
- [8:45 Lo-fi](https://www.youtube.com/watch?v=L1oYdPxGlGA&t=525s)
- [10:03 Nylon EP](https://www.youtube.com/watch?v=L1oYdPxGlGA&t=603s)

## What ships

Forty-four presets. Only two of them are versions of one patch, and they are the place to start: **Just Saw** and **Equal Saw** are identical except for [TUNE](https://benjamindehli.github.io/overtonium/tuning/), and they sound nothing alike.

Each clip plays the same thing: a spread C major chord, played a note at a time and released all at once. This makes it easy to listen through them in order. The clips are rendered by the plugin, not recorded, so they always match what it does. The levels are the ones each preset sets, not matched to each other. Nothing downloads until you press play.

Thirty-nine of them use one of the twelve [oscillator characters](https://benjamindehli.github.io/overtonium/effects/#character). The character is part of the patch like everything else here, so what you hear is the preset as it ships. Keep two of the twelve in mind while listening. Bulb is only audible when the pitch moves, and Op-amp only affects partials above a kilohertz. Both do more when you play them on a keyboard than in a clip of a held chord.

**2-bit Fuzz Organ**

An organ at 8 kHz and 2 bits, where the aliasing and the quantisation noise provide the distortion. The modulation wheel or aftertouch blends the root note towards a fifth

**60s Organ**

A worn combo organ, with a few milliseconds of key click on every note. It goes into a bucket brigade delay driven so hard that the repeats grow instead of fading, then into a spring reverb long enough to still ring when the next chord starts

**6581 Pulse**

The pulse wave of a MOS 6581, the SID chip in a Commodore 64, at half its width, which makes it a square wave. It is on Op-amp, which hardens the partials above a kilohertz

**6581 Sawtooth**

The sawtooth wave of the same chip, on the 6581 character. The character is measured from a real chip, and gives every partial the uneven steps of the chip's converter, the slight bow after it, and its hiss

**6581 Triangle**

The triangle wave of the same chip, also on the 6581 character. A triangle wave has odd harmonics that fall with the square of their number, and the chip's also has a few faint even ones

**8-bit Choir**

A breathy choir on the 6581 character, at 16 kHz and 8 bits. It swells in and out slowly, the breath is blown into [the string in the noise channel](https://benjamindehli.github.io/overtonium/controls/#noise), and each note glides into the next. It goes out through the console

**Big Saw**

A sawtooth with a little randomness in it, so it drifts and beats instead of staying still

**Cathedral**

A principal chorus: octaves and fifths only, a registration an organ builder would recognise. It fades in slowly, the way pipes start to sound

**Chapel**

A small pipe organ in a chapel, on the PD character. The fundamental carries it over a quiet chorus of octaves and fifths, every pipe wavers a little on its wind, and each note starts with a breath. Velocity does nothing, as on an organ, and aftertouch brings the chorus up. It goes into a short digital echo, the room and the resonator's open strings

**Combo Organ**

A transistor combo organ on the Combo character, with one vibrato for the whole keyboard, so every held key wavers together. The modulation wheel or aftertouch brings in the upper partials, and the envelopes and the noise channel give each key a click. It goes into a bucket brigade echo and a spring reverb

**DigiLog**

A digital instrument played through worn analogue gear, at 16 kHz and 8 bits. The envelope delay gets longer for higher partials, so the series comes in from the bottom up. It opens in steps like a filter, not in a smooth sweep

**Dire Dire EP**

A soft, round electric piano of the kind used in underwater music. It is dark except for the 13th and 15th partials, which stand out from their neighbours as the tine

**Drawbar Organ**

Nine drawbars on Wheel, with the key-off stage used as a release click. The upper drawbars jump to 55% for two milliseconds as the contacts break

**Dream Phase**

A strike that turns into a pad, without using separate layers. The short partials make the transient and the long ones come in behind them as it opens. Two partials step on square waves underneath, and the echo is nearly all feedback

**EP Chimes**

A round electric piano with bright chimes above it, at 16 kHz and 8 bits, so the artefacts are part of the sound and not added on top. It goes out through the cassette recorder

**Equal Saw**

Just Saw with every partial moved to equal temperament

**FM Piano**

The electric piano sound of an FM synthesizer such as the DX7, not the instrument it imitates, on FM

**Glass Armonica**

Wet fingers on glass. Each partial arrives later and vibrates faster than the one below it

**Glide**

A pad that comes in from the bottom up. The fundamental starts at once and every partial above it glides in a little later. Played softly it is round and slow, and played hard it is bright and punchy

**Glockenspiel**

A struck bar. Nothing sustains, so every partial fades from the moment it is hit, each over its own length of time. The tone comes from the fourth and eleventh partials more than from the fundamental

**Init**

The neutral starting point: the first three partials and nothing else. It clears the patch and does not change the session settings

**Just Saw**

The same sound as Equal Saw with every partial at its exact whole-number ratio

**KarplusStrong**

A single fundamental and the noise channel's Karplus-Strong string, struck together, with the resonator's piano strings ringing in sympathy, printed to the cassette recorder through the console

**Lo-fi**

8 kHz and 8 bits, soft, with very short partials starting at different times. Close to the sound of a Roland D-50, a Korg Wavestation or a granular synthesizer

**Metallic Piano**

A thin, metallic upright piano. The [resonator](https://benjamindehli.github.io/overtonium/effects/#resonator) rings the strings of the keys you hold, and every string while the sustain pedal is down

**Music Box**

Plucked steel teeth, with irregular vibrato on each partial and one partial bent much further than the rest

**Nylon EP**

A very round electric piano on PD, as if the tines were nylon instead of steel

**Odd Harmonics**

Every even partial is silent, which gives the hollow tone of a stopped pipe. Velocity opens the top of the series

**Omni-84**

The SonicStrings Voice 2 from a Suzuki Omnichord OM-84 System Two

**Pendulum**

Every partial has its own tremolo and pitch movement, faster the higher it sits, and the fast ones die away first, so the movement slows as the note rings, like a pendulum coming to rest. It is on the 6581 character, with the partials spread left and right

**Shimmer**

Every partial rises and falls in level at its own rate, so the spectrum never repeats

**Slow Pad**

Staggered envelope delays, so the spectrum builds up gradually instead of starting at once, with mirrored pans across the series

**Soft Trio**

A Karplus-Strong string from the noise channel, plucked like a clavichord or a guitar, over the first two partials on the Wheel character. Pressure or the mod wheel brings in a vibraphone, with faint partials behind it that swell slowly and wander in pitch. The clip is played without it. Everything sits hard right, and a short digital echo answers on the left at the same level, which doubles it into stereo

**Space Cowboy**

A synthetic tremolo guitar, plucked from [the string in the noise channel](https://benjamindehli.github.io/overtonium/controls/#noise), over a pad with square-wave pitch modulation in some of its partials, and a whistle like a theremin that glides between overlapping notes. The [resonator](https://benjamindehli.github.io/overtonium/effects/#resonator) rings under all of it, and it goes out through the console

**Space Flute**

A pan flute in the distance. The breath is on the noise channel, and the echo and the reverb give the distance

**Sparkle Pad**

Two instruments, split by where a partial sits in the series. The bottom three partials sustain and take up to three quarters of a second to come in. Everything above them starts at once and is gone within half a second

**Stepped**

A slow pad that keeps moving. Seven partials trill on square waves set to whole intervals, two at a fifth and five at an octave. The top three take their level from sample and hold

**Struck Bell**

Higher partials fade first, which is what makes something sound struck. Play harder and the upper partials come in

**StyloPoly**

Voice 3 of a Dubreq Stylophone S-1 with the vibrato on. Unlike the Stylophone it is polyphonic, but the vibrato is shared across the keyboard, as it is on the original with its single oscillator. Every partial's level moves on its own smooth random path

**Synth Ensemble**

Strings and brass in one patch. Nothing in it starts at the same moment, and how hard you play sets how far apart things start. A hard chord starts at once. A quiet one swells in over a couple of seconds, with the series drifting apart on the way

**Tape Choir**

Like a vowel: a formant peak in the low partials instead of a straight roll-off, and an old echo machine with dark, unsteady repeats

**Tape Pluck**

A plucked string on the VCO character, over a quiet pad that swells in behind it and steps in level. The pluck's tremolo is shared across the keyboard, so a chord pulses together. The [resonator](https://benjamindehli.github.io/overtonium/effects/#resonator) rings under it, and it is printed to the cassette recorder through the console

**Vibraphone**

A fast strike where the fundamental rings for seconds and everything above it is gone within one second

**Wurli**

A Wurlitzer 200A electric piano, with one tremolo circuit for the whole keyboard instead of one per tine, so a whole chord pulses together. It goes out through the console

## What a preset carries

The patch and nothing else. Everything in the Settings menu stays as you set it. This is a general rule, not a list that could get out of date: you set up the instrument in the menu and make the sound on the panel. So loading a preset never changes your polyphony, your temperament, your reference pitch, your pitch bend range, your MPE or aftertouch routing, or Adaptive's Retune, harmonic sevenths and No drift. It also leaves your window size and your [LINK](https://benjamindehli.github.io/overtonium/controls/#link) settings alone. These belong to the session you are working in, not to any one sound in it, and a patch that reset you to equal temperament in the middle of a piece would get in the way. The master fader and the output stage are different. They are on the panel and part of the sound, so a preset loads at the level it was made at, through the clipper it was made with.

Presets you save follow the same rule as the ones that ship, and store exactly the same set of controls. If a preset file contains any of these settings, it still loads, and those settings are ignored.

This includes Init. It is a preset too, so it clears the patch and leaves the session settings as you set them.

Values are stored as plain values, not normalised, so a preset still works if a parameter's range is widened later. Anything a file does not mention keeps its default value.

## Switching them over MIDI

A MIDI program change loads a factory preset, so a keyboard, a foot controller or a clip in your host can change the sound without anyone using the menu. The number selects the entry by its position in the Factory list, counting from zero. The list is alphabetical. Hosts and hardware that number program changes from one are one higher, so their 1 is the first entry. A number past the last preset is ignored. It does not wrap round to the start.

The preset is loaded even if it is the one already showing, as a program change does on a hardware instrument with a panel. So a clip that starts with a program change sounds the same every time it plays, and anything you changed by hand is replaced. If you want to keep those edits, save them as your own preset first.

Only the factory presets have numbers. Your own presets live in a folder you can add to and rename at any time, so their positions would not stay fixed long enough for a clip to rely on them. The same applies across versions. A release that adds a factory preset moves every preset that comes after it alphabetically, so a clip made with an older version can select a different sound in a newer one. Sessions are not affected, because a saved session stores the preset's name as well as its number, and the name is what is used.

Program changes work on any channel, MPE included, and load the same things as loading a preset by hand. Your polyphony, temperament and the rest of the Settings menu stay as you set them.

## Saving your own

Saving writes one small file per preset. The menu reads that folder every time it opens, so a preset added from outside the plugin shows up without a restart. The folder is created when it is first needed.

**The bar shows the loaded preset's name on a display of nine cells.** A longer name is shortened in steps, not just cut off. The spaces go first, then the vowels from the right, and only then is the rest trimmed. So *Tape Choir* keeps every letter as TAPECHOIR and *Glass Armonica* becomes GLASSRMNC. The menu, the file on disk and your host all get the full name. Saving does not change the name.

| Platform | Where they live |
|---|---|
| macOS | `~/Library/Application Support/Dehli Musikk/Overtonium/Presets` |
| Windows | `%APPDATA%\Dehli Musikk\Overtonium\Presets` |
| Linux | `~/.config/Dehli Musikk/Overtonium/Presets` |

They are small XML files of parameter values, plus the channels of any macro set to Selected, so you can read them, edit them, keep them under version control and share them by copying the file. The menu sorts them by name, not in the order the filesystem returns them.

**Folders become groups.** Make a folder in the preset folder and it appears in the menu as a submenu with everything inside it, nested as deeply as you nest the folders. There is nothing to set up and nothing else to manage. The folders are the groups, so to move a preset between groups you move the file. Groups are listed before the presets that are not in a group, and the loaded preset has a tick wherever it is.

---

Page: Installing Overtonium on macOS, Windows and Linux
URL: https://benjamindehli.github.io/overtonium/install/
Last changed: 2026-11-01

# Installing Overtonium

Installers for macOS and Windows, and a zip for Linux. Every release also has a zip next to the installer with the same builds unpacked, for anyone without administrator rights or who keeps plugins in their own folders.

## Download

Five files. Most people want the installer for their platform.

| File | Holds | When to use it |
|---|---|---|
| [macOS installer](https://github.com/benjamindehli/overtonium/releases/download/v1.16.0/Overtonium-v1.16.0-macOS.pkg) | VST3, Audio Unit, Standalone | The usual choice on a Mac. You choose which of the three formats to install |
| [macOS zip](https://github.com/benjamindehli/overtonium/releases/download/v1.16.0/Overtonium-v1.16.0-macOS.zip) | VST3, Audio Unit, Standalone | On a Mac without administrator rights, or if you keep your plugins in a folder of your own |
| [Windows installer](https://github.com/benjamindehli/overtonium/releases/download/v1.16.0/Overtonium-v1.16.0-Windows.exe) | VST3, Standalone | The usual choice on Windows. It needs administrator rights, and SmartScreen will say the publisher is unknown |
| [Windows zip](https://github.com/benjamindehli/overtonium/releases/download/v1.16.0/Overtonium-v1.16.0-Windows.zip) | VST3, Standalone | On Windows without administrator rights, if you keep your plugins in a folder of your own, or if you would rather not run an unsigned installer |
| [Linux zip](https://github.com/benjamindehli/overtonium/releases/download/v1.16.0/Overtonium-v1.16.0-Linux.zip) | VST3, LV2, Standalone | The only download for Linux. There is no installer, because Linux distributions do not agree on where plugins go |

Every file above is from **version 1.16.0**, released 1 November 2026. [What changed in it](https://benjamindehli.github.io/overtonium/releases/), and in every version before it, has a page of its own. Older builds and the source are on the [releases page](https://github.com/benjamindehli/overtonium/releases/latest).

## What each platform gets

| Platform | Installer | Formats |
|---|---|---|
| macOS | `.pkg`, with a chooser | VST3, Audio Unit, Standalone |
| Windows | `.exe` | VST3, Standalone |
| Linux | none, use the zip | VST3, LV2, Standalone |

### macOS

The package asks which of the three formats you want and installs only those. It is a universal binary, so one download covers Apple Silicon and Intel, and it is signed and notarised, so it opens without a warning. The minimum is macOS 10.13 on Intel and 11.0 on Apple Silicon, which is the lowest version the toolchain supports for that architecture.

### Windows

The installer is not signed, so SmartScreen says the publisher is unknown. Choose "More info" and then "Run anyway". If you would rather not, [build it from source](https://benjamindehli.github.io/overtonium/install/#building) instead. The VST3 goes to the shared plugin folder, so the installer asks for administrator rights.

### Linux

There is no installer, because Linux distributions do not agree on an install location. The zip holds VST3, LV2 and standalone, and the two plugin folders go in `~/.vst3` and `~/.lv2`.

## When a host does not see it

| Symptom | What to try |
|---|---|
| Logic or GarageBand shows no Overtonium | They only look for new Audio Units when they start. Quit and reopen instead of rescanning |
| A host lists it but will not load it | Check you have the right format. Logic uses the Audio Unit and most others use the VST3. It is easy to install only one of the two by accident in the chooser |
| Nothing appears after installing on macOS | Confirm where the files went with `pkgutil --files com.dehlimusikk.overtonium.vst3`. The plugins belong in `/Library/Audio/Plug-Ins` and the app in `/Applications` |
| The plugin loads but is silent | Check the master fader and whether a channel is [soloed](https://benjamindehli.github.io/overtonium/controls/). Solo is per channel and mute overrides solo, so a soloed strip can still be muted |
| Held notes sound but nothing new does | Overtonium is a synthesizer and needs MIDI. Check the track's input, not the plugin |

Please report anything else on the [issue tracker](https://github.com/benjamindehli/overtonium/issues), especially on Windows and Linux, for the reason below.

## Telling you about new versions

The check is off unless you turn it on. The first time you open the window, a dialog asks whether to turn it on. Nothing goes out to the network unless you say yes. Either answer is remembered, so you are not asked again. After that, the setting is under Settings.

With it on, the plugin downloads one small file from this site when the window opens and compares the version in it with the build you are running. If there is a newer version, a dialog says so, once for each release, SETTINGS lights up, and the first item in its menu opens the release page. It is a plain request for a static file: no identifier, no telemetry, nothing about your machine. The [security policy](https://github.com/benjamindehli/overtonium/blob/main/SECURITY.md) lists exactly what is sent.

The dialogs are drawn inside the plugin's own window, not opened as separate windows. A plugin cannot tell whether someone is opening it or a host is scanning its plugin folder at startup, and a window that waited for an answer would stop the scan. These dialogs do not wait for anything, and a question only counts as asked once someone has answered it.

## What it costs to run

Polyphony is what matters most, because eight voices of 32 partials is 256 sine oscillators. Measured on one core at 48 kHz with everything running at once: both modulators on every channel, drift, velocity, aftertouch, the meters and the noise channel.

| Voices | Oscillators | One core |
|---|---|---|
| 1 | 32 | about 1% |
| 8, the default | 256 | about 7% |
| 16 | 512 | 14 to 15% |

The effects on the bar add to that, and cost the same at any polyphony, because they process the sum and not each voice. Most of them cost far less than the voices: the room and the plate are around half a percent each and the echoes less. The spring is the exception at about two and a half percent, because its dispersion is a chain of filters per spring and there are three springs. The resonator costs about a tenth of what eight voices do in Piano or Octave. In Open, or in Piano with the sustain pedal down, all eighty-eight of its strings are free to ring, and it costs about half of what eight voices do. These are one machine's figures and not a promise about yours, but the proportions hold: the voices are the main cost, and the resonator with all its strings free is the only effect that comes near them.

**The render rate is the control to reach for** if you need it cheaper. Turning it down really lowers the rate, it is not a filter on top. The whole voice pool renders at the lower rate, and at 8 kHz it costs a fifth as much. You can also hear it, which is why it is on the panel and not in a preferences dialog. Lowering the polyphony also works, in proportion.

What keeps it affordable is that the engine stays a plain bank of oscillators. It uses a 4096-point interpolated sine table instead of a sine call per sample. Modulators and envelope coefficients update once per 32-sample block, and only the oscillator and the envelope run at sample rate. A partial that is muted, faded out or at zero skips the oscillator entirely.

## What has been tested

All three platforms are built and run through the test suite on every push, and the VST3 is checked with pluginval at strictness 8 on each of them. macOS is the only platform where the plugin has been loaded into a real host, and there the Audio Unit and the VST3 also pass `auval`. Nobody has run the Windows or Linux builds in a DAW yet, so treat those as untested.

## Building from source

You need CMake 3.22 or newer and a C++17 compiler. JUCE is downloaded at configure time, so there is nothing else to install on macOS or Windows.

```
git clone https://github.com/benjamindehli/overtonium.git
cd overtonium
cmake -B build -DCMAKE_BUILD_TYPE=Release
cmake --build build
```

The plugins are copied into your user plugin folders as part of the build. Pass `-DOVERTONIUM_INSTALL_AFTER_BUILD=OFF` to skip that. Linux needs a handful of development packages first, which the [contributing guide](https://github.com/benjamindehli/overtonium/blob/main/CONTRIBUTING.md#linux) lists.

The build treats warnings as errors, the same as the project's own checks. If your compiler is one the project has not been tested with, add `-DOVERTONIUM_WARNINGS_AS_ERRORS=OFF` to the first cmake command and it builds anyway.

Overtonium is free software under the AGPLv3, which follows from JUCE. The whole source, including the tests and the packaging, is in the [repository](https://github.com/benjamindehli/overtonium).

---

Page: Overtonium releases: what changed in each version
URL: https://benjamindehli.github.io/overtonium/releases/
Last changed: 2026-11-08

# Overtonium releases

What changed in each version, newest first. Each one links to its full notes, which say more and carry the installing and upgrading detail as well.

## 1.17.0, 9 November 2026

**Velocity in decibels and a KEYBOARD row (#105, #106).** A dB switch beside the VELOCITY caption makes equal steps of the knob equal steps of loudness, up to 40 dB on the softest note, and older patches stay on the Classic curve. The KEYBOARD row in OUTPUT voices a partial across the keyboard, tilting it around middle C or fading it out of part of the keyboard.

**Pedals and the resonator (#118 to #122).** The sostenuto pedal, and half-pedalling on a continuous sustain pedal. The resonator's strings follow STRETCH and ring back on the side of the partials that drive them, and two new modes: Fifths, a drone of the root and its fifth, and Chord, which keeps the last chord ringing under a line.

**Smaller changes (#101 to #104).** Option-click or Alt-click on M or S mutes or solos only that channel. The shape menus can set every channel to any shape. Zoom goes to 175% and 200%. A row's name in the gutter says what the row does.

**Three new presets and a quieter panel.** Chapel, Pendulum and Soft Trio, forty-four in all, and the factory presets moved to the dB curve. The meters and macro faders sit in recesses, and the noise channel is marked with a noise trace.

**Fixes.** Switching from Pure or Bulb no longer drops the sound out, two bounces of the same passage come out the same, MIDI Learn replaces a control's controller, the tail counts macros on the release, a session file holding a value that is not a number no longer silences the plugin, and two ways the window could crash the host are closed.

**Before upgrading.** The resonator's mode list grows from five choices to seven and each macro's row list from twenty to twenty-one, so automation drawn before lands a little further along. MIDI program changes from 8 up land later. The factory presets play softer notes differently. 884 parameters.

## 1.16.0, 1 November 2026

**Two characters and three output machines, measured.** VCO is an analogue oscillator's sine and the mixer it feeds, which clips one side first when a chord is played hard. Combo is a transistor combo organ's 8' lever, a filtered square with only odd harmonics, where every partial takes the tone of the key you play. Console is a mixing console's mix bus, a transformer, a compressor and a tape stage. Cassette is a four-track cassette recorder with Dolby B, its saturation, its flutter and its hiss. Print is Console recorded to Cassette, with the hiss 6 dB lower.

**Pick channels by their numbers (#75).** Cmd-click or Ctrl-click a channel's number, or drag across the numbers, and LINK's new Selected scope moves only those channels. A macro can take them too, and keeps its own copy.

**Three new presets.** 8-bit Choir, Combo Organ and Tape Pluck, one on each new character, and four presets moved to the new machines. Forty-one in all.

**Colours, less CPU and fixes.** The characters and the machines on the top bar each light in a colour that is easy to tell apart, and the knobs beside each button light with it. A silenced channel's shape screens switch off. The engine uses about a tenth less CPU, and the output stage almost none unless Console, Cassette or Print is chosen. The tool button shows the chosen tool at once, a preset file holding a value that is not a number no longer silences the plugin, and the update notice only opens the project's own pages.

**Before upgrading.** CHARACTER grows from ten choices to twelve, the clip type from five to eight and each macro's scope from fifteen to sixteen, so automation drawn before lands a little further along. MIDI program changes from 5 up land later. Four presets sound different. Still 850 parameters.

[The full notes for 1.16.0](https://github.com/benjamindehli/overtonium/releases/tag/v1.16.0)

## 1.15.0, 25 October 2026

**Four characters measured from real instruments.** Wheel is a tonewheel organ's generator, with each wheel's own harmonics, its wobble as it turns and the faint leakage of the wheels near each note. FM and PD are the nearly pure sines of a frequency modulation and a phase distortion synthesizer, with the noise measured under their notes, and on PD the sum and difference tones its mix gives a chord. 6581 is a sine as a 6581 sound chip would play it, through its uneven 12-bit converter, with its bow, its sum and difference tones and its hiss. FM, PD and 6581 light the character button in their instrument's colour.

**Kirnberger III, Vallotti and the wolf on G#-Eb.** Two more temperaments, nine in all. Meantone and Pythagorean now put their wolf on G#-Eb, as organs and harpsichords do, which keeps C, F, B flat, G, D and A major in tune on C.

**Presets and fixes.** 6581 Pulse, 6581 Sawtooth and KarplusStrong are new, thirty-eight in all, and 6581 Triangle, Drawbar Organ, FM Piano and Nylon EP moved to the new characters. The bus drive no longer loses about a decibel of the top octave, and no longer pops on loud notes on Valve, Rail and Diode. The lamps follow the note you just played.

**Before upgrading.** CHARACTER grows from six choices to ten and the temperament list from seven to nine, so automation drawn before lands a little further along. Sessions saved with meantone or Pythagorean put the wolf in a new place. A MIDI program change from 2 up lands two presets later, and from 17 up three later, while a saved session keeps its preset by name. Still 850 parameters.

[The full notes for 1.15.0](https://github.com/benjamindehli/overtonium/releases/tag/v1.15.0)

## 1.14.0, 16 October 2026

**A string in the noise channel.** A Karplus-Strong string the noise strikes, tuned to the note and following the wheel, glide, MPE and every temperament. STRING, DECAY, DAMP and POINT sit under COLOUR. A short noise envelope plucks it and a held one blows into it, so a pluck and a pitched breath are settings, not modes.

**The resonator.** Eighty-eight sympathetic strings that ring with what you play, before the bus drive, in their own frame on the top bar. Piano frees the strings of held keys and the sustain pedal frees the rest, Open keeps the pedal down, Octave rings twelve strings from C3 and Series the root's own harmonics. MIX, DECAY and DAMP keep the level about the same across their range. The new preset Space Cowboy uses both, and Metallic Piano now rings its own strings.

**A rearranged top bar and a faster mixer.** The four buttons come first, then the wordmark over the preset screen, with the spacing evened out. The update question is asked inside the window, and SETTINGS stays lit while a new version waits. Silenced channels and switched-off stages turn their lights fully off. Soloing, scrolling and resizing redraw the mixer far faster. A wheel over a knob in the bars no longer scrolls the mixer, Fit all 32 channels gives every row its full height, and JUCE 9.0.4 fixes a possible crash in Windows hosts when a window closes.

**Before upgrading.** Space Cowboy sorts at program 25, so a MIDI program change from 25 up lands one preset later, while a saved session keeps its preset by name. Metallic Piano sounds fuller than it did. There are 850 parameters, the eight new ones on the end.

[The full notes for 1.14.0](https://github.com/benjamindehli/overtonium/releases/tag/v1.14.0)

## 1.13.0, 13 October 2026

**Adaptive, a temperament that retunes as you play.** After Eivind Groven's organs, every pitch class has five versions a comma apart, and the chord you hold is kept on whichever make it purest, settled again every time a key goes down or comes up. A held note the next chord needs elsewhere slides there over the Retune time. A dominant seventh is tuned on the harmonic seventh, 4:5:6:7, and NO DRIFT keeps the keyboard home where a progression would carry it off a comma at a time. Every one of the 125 chord types in music-utils is played through it in all twelve keys on every build.

**Two bars under the top bar.** PLAYING and TUNING open a bar each, both closed until you open them. The Tuning bar shows twelve displays, one for each note, saying how far it sits from equal temperament in cents, in every temperament, and under Adaptive a PURE lamp, 7:4, DRIFT, NO DRIFT and the Retune knob. The Playing bar holds the voice and expression settings, shows the voices in use, the sustain pedal, the pitch wheel, the mod wheel and pressure, and carries the eight macros as knobs captioned with what they are aimed at, which a right-click assigns, opens in MACROS or binds to a controller.

**Displays that behave like real ones, and No stack.** Every segment display is a fixed row of cells that light or stay dark, with no dimmed figures, cleaner letters and a proper sharp sign. One voice per key is renamed No stack, for what it does. Right-clicking a knob on the top bar now offers MIDI Learn, which it never did.

**Before upgrading.** Adaptive is a seventh entry on the temperament list, so temperament automation drawn against six lands a little further along, while a saved session keeps its temperament. A host shows No Stack, Pitch Mod Shared and Amp Mod Shared in place of the old names, with the same ids. There are 842 parameters.

[The full notes for 1.13.0](https://github.com/benjamindehli/overtonium/releases/tag/v1.13.0)

## 1.12.0, 10 October 2026

**Glide on every channel.** Every channel has its own glide time, the last row under PITCH MOD, from Off to five seconds. Because each partial travels on its own time, the series does not have to arrive together: short times on the top land the brightness first and let the pitch follow, and the other way round the note arrives and its colour catches up. The needle on each partial's PITCH MOD rule follows the glide back to centre. Two switches govern every channel at once, as lit letters beside the GLIDE caption: **L** glides always or only legato, and **T** makes the time a fixed rate per octave or a fixed time however far a note goes.

**ATTACK reaches into where a partial starts.** The PHASE knob is gone. Turned past its shortest time, ATTACK reads in degrees, from a rising zero crossing at 0° to the partial's own peak at 90°, and STRIKE moves along the whole knob, so a hard blow lands on the peak and a light one travels up into a short attack.

**Legato is a switch, and with more than one voice it plays a top line.** With one voice it is the monophonic legato it always was. With more, a melody held from note to note carries on over chords that still retrigger each time their keys come up and go down, and the note a melody carries on from always sounds. A wrong guess about which key is the melody costs an attack and nothing more.

**Shared by all notes, a thirty-fourth preset, and a quieter panel.** The in-phase switch is renamed for what it does and has a switch of its own beside each SHAPE caption. *Glide* is a pad whose partials glide in up the series, round played softly and bright played hard. The panel is a shade darker with lighter captions, which now clear the 4.5:1 contrast the accessibility standard asks for, and a dark sliver at the top left corner of every screen and lamp is gone.

**Before upgrading.** The Limiter's lookahead is off by default, so new instances report 12 samples of latency at 48 kHz rather than 108, and a saved session keeps its setting. Start phase is gone, so a patch that set one starts each partial at the zero crossing, and an automation lane on it in a host that goes by position now moves glide. ATTACK automation drawn against the old travel lands on different times. *Glide* sorts after *Glass Armonica*, so a MIDI program change will name the preset above the one it used to, while a saved session carries the name and is unaffected. There are 839 parameters.

[The full notes for 1.12.0](https://github.com/benjamindehli/overtonium/releases/tag/v1.12.0)

## 1.11.1, 7 October 2026

**Four controls read to seven decimal places.** The pitch modulator's rate and depth, the drift and the amplitude modulator's rate had no text of their own, so a host's automation lane, the panel's value popup and a screen reader all said `4.0000005 Hz` for four. They now read as the other controls do, `4.00 Hz`, `30.0 Hz` and `0 ct`. The macro amount reads as the panel shows it, `-100.0 %` to `+100.0 %`, rather than to four places, and a value typed into a host is read as a percentage. Only the text changed: ranges, defaults and stored values are as they were, so sessions, presets and automation play back exactly as before.

[The full notes for 1.11.1](https://github.com/benjamindehli/overtonium/releases/tag/v1.11.1)

## 1.11.0, 7 October 2026

**Eight macros, so a relationship across the series can be automated.** A LINK drag moves 32 parameters and a host catches only the last one touched, so a relationship you can edit by hand could not be recorded at all. A macro names a row, a scope and a curve, and gives that relationship one control and the host one lane. It offsets rather than writes, so the patch underneath is untouched, which is why the knobs do not move when a macro does. A driven control says both things at once instead: the ring of ticks lights to where the macro has taken the value in the macro's colour, and the pointer stays at what the patch holds. A fader says it in the gutters either side of its track. Eight is a pool rather than a count, since a plugin declares its parameters once and cannot add one later.

**MIDI Learn.** Right-click a control, choose MIDI Learn, move something on the controller. One parameter per controller number, saved with the session rather than with a patch, because which controller moves which knob describes the desk rather than the sound. Eleven numbers are refused rather than bound, the mod wheel and the MPE slide among them, so arming a knob and then reaching for the slide to steady a note leaves the note steadied and the knob still waiting.

**CLIP is five machines instead of a switch.** *Soft* is the curve that was already there and came through bit for bit. *Hard* is a wall at unity. *Asymmetric* lets one half of the wave lean over first, the way a single-ended stage runs out of supply in one direction, which gives even harmonics where the others give odd. *Limiter* does not bend the wave at all, it moves the level. *Fold* turns the wave back on itself and sings where the rest crunch. All five are bounded by full scale, tested at drives up to 400 times at three sample rates. The limiter looks ahead two milliseconds, which is what took the reported latency from 12 samples to 108 at 48 kHz, and Settings can refuse it.

**The parameters scroll.** The rows are small on a 1080p screen and zooming in to read them made the window taller than such a screen has room for. A channel is three bands now: the channel number at the top and the fader at the foot are pinned, and everything between them scrolls behind an opening, so neither of the two things you find your place by is ever what has scrolled away.

**The panel was rebuilt.** Every switch is a lit plastic cap in a moulded well, with a fixed near-black legend printed on the plastic so a word does not change colour under its own lamp. The knobs are discs with an eight-slot collar, and the notches are cut out of it rather than painted on. The preset name is spelled on a fourteen-bar display. The OUTPUT rule carries two lamps, VELOCITY on the left and AFTERTOUCH on the right, so all five rules say something. LINK and DRAW became one tool button rather than two switches that could not both be on, and choosing a scope now lights the channel number of every strip a drag would reach.

**A thirty-third preset, and an accessibility statement.** *6581 Triangle* is the triangle of a MOS 6581, the SID in a Commodore 64, which is not quite a triangle: the chip draws its ramp as a staircase of 2048 steps, so it carries even harmonics a triangle should not have. `ACCESSIBILITY.md` says what is checked rather than intended, and says plainly that the plugin's own window is not keyboard operable and that nobody who relies on assistive technology has evaluated any of it.

**Fixed.** A partial silenced by a negative AFTERTOUCH amount took every lamp on its channel dark with it, which had been wrong for the envelope and tremolo lamps since those lamps existed. And a note played while the channel was already under pressure thumped or faded in, because the pressure smoother started from nothing at every note-on instead of from where the channel was holding it.

**Before upgrading.** The plugin reports 108 samples of latency at 48 kHz where it reported 12, which is the limiter's window, and Settings can take it back. Loading a preset now moves the master fader and the clipper, since both left the settings a preset cannot reach. *6581 Triangle* sorts third, so a MIDI program change will name the preset above the one it used to, while a saved session carries the name as well as the number and is unaffected. Fifty parameters were added, 836 in all, all of them on the end, so every automation lane written before this release still points at the control it was pointing at.

[The full notes for 1.11.0](https://github.com/benjamindehli/overtonium/releases/tag/v1.11.0)

## 1.10.1, 30 September 2026

**Removing the plugin could hang the host.** In a host that unloads the plugin binary once no instances are left, taking the last one out could lock the host up, with no way out but to end the process. Reported in FL Studio and in Reaper, while hosts that keep the binary loaded were never affected.

The texture the panels are drawn with is an image, built the first time a window opens and shared by every window after it, and it was being kept for the life of the process. Under Windows that image lives on the graphics card, so letting go of it hands resources back to the driver, and keeping it in a static meant that happened while the host was unloading the binary. Windows holds a lock for the whole of an unload and the driver work cannot run until that lock is free, so neither finishes. The texture now belongs to the windows that use it and goes when the last one closes. Nothing else changed, and nothing about the sound has.

## 1.10.0, 30 September 2026

**The echo is three machines, and so is the reverb.** Both were a switch that turned one thing on, and both are now a button that says which machine is running. *Tape* is the loop that was already there. *BBD* is a line of buckets, which gets darker the longer you set it to, because a longer delay can only be had by clocking the buckets more slowly and the filter has to come down with the clock. *Digital* does not pretend to be anything and is the only ping-pong of the three, crossing every repeat to the other side where the other two never cross at all. AGE is the same knob on all three and asks each for the thing that machine gets worse at: tape wear, a slowing clock and a breathing compander, or falling bits and sample rate.

*Room* is the reverb that was already there. *Modulated plate* is a sheet of steel with no geometry to hear, so a hit spreads across all of it at once and the tail is dense from its first instant rather than arriving in stages. It is called modulated because it is: its allpasses wander six times as far as the shape they come from gives them, and without that a long decay repeats its own pattern instead of washing. *Spring* is a tray of three helical springs, and a helix carries the top of the band faster than the bottom, so a hit comes back as a chirp that starts high and falls and every bounce adds another. DECAY, DAMP and PRE mean the same thing on all three, and the wet levels are matched, so switching machines changes the character of a tail rather than its length or its loudness.

**The character now says how hard the mix is driven.** The thirty-three channels are summed onto a bus that drives, by an amount belonging to the oscillator character. It answers to level rather than to a setting, which is what a table baked into one partial's waveform cannot do: play quietly and it is barely there, lean on the keyboard and it arrives. Bulb asks for the most and Diode the least, Pure sums cleanly and does nothing, and there is no knob, because how hard a summing amplifier is driven is part of what choosing a character means.

**How fast you let a key go sets the level its tail starts from.** On a keyboard that senses a release, the hardest lift doubles the key-off level and the softest halves it. There is no knob and no parameter: it is a property of the gesture the way velocity is, and a keyboard that cannot sense a release changes nothing.

**The master volume is a glass fader lying across the output meter**, over the lamps it sets the level of, instead of a knob in a group of its own. A LINK drag that spreads a move uniformly now spreads it in decibels as well: on the fader row it was uniform in travel, which put six or seven decibels on the loudest channel and thirty on the quietest.

**A thirty-second factory preset, and four others that reach for the new machines.** *60s Organ* is a combo organ that has not been serviced: the partials are pulled off each other rather than tuned, so the stack beats against itself instead of fusing, and the noise channel is a few milliseconds of key click on every note. It runs into a bucket brigade driven hard enough that the repeats grow rather than fade, and falls into a spring still ringing when the next chord arrives. It is also the first patch to ask for Diode. *Shimmer* and *Tape Choir* now ask for the plate rather than the room, and *Cathedral* and *Struck Bell* stay on the room. Because the list is alphabetical and 60s Organ sorts second, a MIDI program change now names the preset above the one it used to. A saved session is unaffected, since it carries the name as well as the number.

**LINK came back switched off.** Its scope and its curve were remembered with the session and the switch that decides whether either applies was not, so reopening a window gave you exactly the drag you had set up, turned off. All three are saved together now.

**Unfolding a section made the faders swell.** On a window shrunk until the faders were squeezed, unfolding gave back twice the height it had taken, all of it to the faders. Applying the window's new size limits is itself a resize, so the height was measured after it had already moved and the same rows counted twice.

**Draw the faders.** A drag across the fader area sets every channel it passes over from the pointer's height, so a spectrum goes in with one stroke instead of thirty-two drags. Hold shift for it, or latch it on with the new DRAW switch under the LEVEL caption. Armed, every fader lights in the same accent the switch does, so what a drag would reach is visible before you touch anything. It is not LINK under another name: LINK shares one relative move out across the channels by a rule, and this sets absolute values freehand. Columns crossed between one moment and the next are filled in, the whole stroke is one undo, and the noise channel draws with the rest. While it is armed the pointer becomes a pencil and LINK reads as off, without being off.

**A section folds from any strip, not just from the gutter.** Every section draws a rule across all thirty-three columns, lined up with its heading in the caption gutter on the far left, and clicking one of those rules now folds that section. Reaching the gutter meant leaving the partial you were working on, crossing the mixer and finding your way back, by which time you had lost which column you were in. The pointer shows the same hand over a rule as over a heading.

**The bar says what each group is.** A small caption under the button at the head of the character, echo and reverb groups, because those buttons show a value and nothing said what OP-AMP, TAPE or ROOM was a choice of. With the caption doing that, the button reads OFF when the effect is off, where it used to read ECHO or REVERB so the group could still be found, and the ring of lit ticks around that effect's four knobs goes dark with it. The knobs keep their positions, so a setting can be dialled in before the thing is switched on. The clipper also has a switch on the bar now, CLIP, under the output meter beside the two converter readouts, which are a little larger themselves. It is still in the Settings menu too.

**Fixed.** Right-clicking the gap between two sections opened two LINK menus stacked on each other, and picking an item on the front one left the one behind it unchanged, so the setting looked to have been refused when it had already taken. A scroll over a knob also dragged the mixer sideways whenever the window was narrow enough for the series to need scrolling, since each strip watches everything inside it and passed the wheel on after the knob had taken it. A scroll on a strip's background still moves the series. The plugin would not tell you which version it was: not in the Settings menu, not on the panel, not anywhere, while the bug report template refused a report without one and said to read it off that menu. It is the menu's first heading now. The first repeat of an echo came back clean while every repeat after it was worn, on both the tape and the bucket brigade. Both echoes compressed everything they were handed, because the tape's backstop was set for a bus running near one where a full patch runs closer to two. The digital delay had a tail that never went away at any feedback from a half upwards, which is what a quantiser inside a feedback loop does. And the zoom menu ticked 100% whatever the zoom was.

**Before upgrading.** Presets on a character sound fuller, because the bus they are summed onto now drives, and that is the intended sound of them. The echoes are cleaner and their first repeat is coloured, so a patch leaning on the tape echo's compression will not find it. *Sparkle Pad* is louder and otherwise unchanged, since it was mixed low on purpose to keep that compression down. Your own patches get the tape and the room, which is what every patch written before this had. Two parameters were added, 786 in all, both on the end, so every automation lane written before this release still points at the control it was pointing at.

[The full notes for 1.10.0](https://github.com/benjamindehli/overtonium/releases/tag/v1.10.0)

## 1.9.0, 23 September 2026

**Which oscillator the partials are.** A new control at the head of the series group, offering six of them. All six are sine oscillators, and what differs is the way each circuit fails to make one. *Pure* is the sine every partial already was. *Bulb* is a Wien bridge whose lamp lags, so the level sags behind every move of the pitch and recovers over about half a second. *Rail* grows into its own supply, *Diode* is bent into shape by two of them that do not match, and *Valve* is a triode with an octave on top of every partial. *Op-amp* is an amplifier that cannot keep up, so the top of the series hardens as it climbs while the bottom stays where it was. Each one is also a rack of thirty-two rather than one oscillator used thirty-two times: a fixed spread of a couple of cents and a fraction of a decibel, the same every time you load the patch. It is part of the patch, so a preset carries it, and twenty-five of the thirty-one factory presets now ask for one: eleven on Bulb, eight on Op-amp, four on Valve and two on Rail. Loading one of those gives what it ships as rather than what it was in 1.8.0, and setting the character back to Pure gives the older sound, except on *Synth Ensemble*, which was reworked further: its velocity amounts now take more of the series, so a soft chord arrives with less above the first few partials, and its reverb sits a third in rather than a quarter. A session is unaffected either way, since it carries the values rather than the preset's name.

**A MIDI program change loads a factory preset**, by its position in the alphabetical Factory list counting from zero, so a keyboard, a foot controller or a clip can change the sound without anyone touching the menu. A number past the last preset is ignored rather than wrapping.

**The converter readouts say what they are counting**, reading `48.0 kHz` and `8 bit` where they used to have room for the figure alone. LINK moved to the top of the caption column on the left to make that room, over the rows it gangs, and with it gone the bar fits on one row at the width the window opens at. Every switch on the panel now lights its word rather than its face, every screen and lamp is set into it with the lip of the opening it sits in, the standalone wears the instrument's own title bar instead of the grey-green one every unstyled JUCE application gets, and Settings has an entry that puts a window dragged narrow back to the width that shows all 32 channels.

**A modulator can be one circuit the whole keyboard hears.** Both belong to the note by default, so a chord struck a note at a time gives you three tremolos breathing out of step. The new entry at the foot of a shape button's menu, one for each modulator, hands the phase to the channel instead, so a note arriving late joins what is already running and a chord breathes as one thing. *Wurli* ships sharing its tremolo and *StyloPoly* its vibrato.

**Undo took a step several times a second while you were still turning a knob**, which was most obvious with a scroll wheel. A gesture is one step now, however many values it moved and however long it went on, and a gesture that ends where it began is not a step at all. A host playing an automation lane used to fill the history as well, and no longer leaves any: the history holds what a person did.

**Before upgrading.** Nothing you have made moves: a session opens sounding as it did, and your own presets get Pure and a modulator per note, which is what every patch written before this had. Twenty-five of the factory presets do sound different, since they now ask for a character, and two of those share a modulator across the keyboard as well. Three parameters were added, 784 in all, all three on the end, so every automation lane written before this release still points at the control it was pointing at.

[The full notes for 1.9.0](https://github.com/benjamindehli/overtonium/releases/tag/v1.9.0)

## 1.8.0, 19 September 2026

**Three more factory presets, thirty-one in all.** All three lean on STRIKE, which arrived in 1.7.0 and is what lets one patch play two ways depending on how it is struck. *Glockenspiel* is a struck bar whose tone sits in the fourth and eleventh partials rather than the fundamental, with nothing sustaining and every partial ringing down for its own length. *Dream Phase* is a strike that becomes a pad, the short partials the transient and the long ones arriving behind them. *Sparkle Pad* splits the series in two: the bottom three sustain and arrive slowly, everything above attacks at once and is gone inside half a second.

**A session could open with the wrong preset showing as selected.** Presets sort alphabetically, so adding one moved every preset after it and an index saved last year named a different sound. The sound was never in doubt, since a session stores its values, but a host's program menu would highlight a patch the session was not playing. The saved name decides now, so reordering presets can no longer move a session onto a different one.

**Built on JUCE 9.0.2**, up from 8.0.15, with nothing in Overtonium needing to change for it. Every audio example on this site renders to the same samples before and after. Building from source on Linux now also needs `libxi-dev` and `libegl-dev`.

**Before upgrading.** The factory presets are renumbered, since the three new ones sort into alphabetical place. The sound is stored with the session either way, and with the fix above the name you saved is what the menu goes by. No parameter was added or removed, so automation is unaffected.

[The full notes for 1.8.0](https://github.com/benjamindehli/overtonium/releases/tag/v1.8.0)

## 1.7.0, 12 September 2026

**STRIKE aims key velocity at the front of the envelope.** It sits at the head of the ENVELOPE group on every one of the thirty-three channels, above DELAY and ATTACK, because it is about the blow rather than about the shape. A hard note starts sooner and arrives faster, a soft one waits and then opens slowly. Each knob below it stays the limit of its own row: velocity only ever lengthens the attack and only ever pulls the delay in, so ATTACK goes on meaning the fastest that partial gets and DELAY the latest it ever arrives. The range is wide because the attacks worth stretching are short, so 2 ms under a hard blow is half a second under the softest. *Synth Ensemble* is rebuilt around it.

**MPE slide no longer lurches** on a controller whose slide axis rests at one end rather than at the centre. It is read as movement now: the first value a note is given becomes that note's nought, and what reaches the sound is how far the axis has travelled from it since. A Seaboard behaves as it always did. An Expressive E Osmose, which spends the first part of the key travel on pressure and only then starts sending slide, no longer jumps darker the instant the axis engages.

**Before upgrading.** LIFT is gone. It aimed release velocity at the key-off level and was the one control here that most keyboards cannot play, so a patch that used it loses those values and an automation lane pointed at one finds nothing. The parameter count is unchanged at 781 and everything else is untouched: every factory preset but *Synth Ensemble* renders to a byte-identical file before and after. A finger landing forward on a pad no longer starts the note bright, which is what reading slide as movement costs.

[The full notes for 1.7.0](https://github.com/benjamindehli/overtonium/releases/tag/v1.7.0)

## 1.6.1, 11 September 2026

One fix, and nothing else. The waveform selectors added in 1.6.0 kept listening to their parameter fractionally longer than they should have, into the moment they were being taken apart, so a value arriving at exactly that instant landed on a control that was already half gone. Closing a plugin window while automation plays is the everyday version of that, which is why it was worth a release of its own rather than waiting.

[The full notes for 1.6.1](https://github.com/benjamindehli/overtonium/releases/tag/v1.6.1)

## 1.6.0, 10 September 2026

**Both modulators pick a waveform.** Eight shapes on pitch and seven on amplitude, on every one of the thirty-three channels, so the fifth partial can be gated while the fourth breathes. The pitch depth reaches an octave now, which is what the shapes that step rather than sweep are worth having: a square on the pitch is a trill and sample and hold is a run of random notes. The tuning digits and the shape glyph are lit, so a strip has two displays rather than a screen and a label.

Two more factory presets, *Stepped* and *StyloPoly*, twenty-eight in all. Loading a preset used to tell the host about a thousand parameter changes, most of which were not changes, and now tells it only what moved.

**Before upgrading.** The factory presets are renumbered, since the two new ones sort into alphabetical place rather than onto the end, so a session that stored a preset by number opens on a different one. The sound is stored with the session either way. Automation is unaffected: the 65 new parameters are inserted among the existing ones, and both formats identify a parameter by name rather than by position.

[The full notes for 1.6.0](https://github.com/benjamindehli/overtonium/releases/tag/v1.6.0)

## 1.5.0, 5 September 2026

Seven more factory presets, twenty-six in all, six of them the same electric piano seen from different angles, because that is what a stack of sine waves is good at. The preset menu became two submenus rather than one long list, and folders in your preset folder became groups in the menu.

**Nothing in the Settings menu travels with a preset any more.** Saving one used to write your polyphony, temperament, bend range and MPE into the file, so recalling a sound reached over and changed how the instrument was set up. The factory presets already knew better, and both kinds follow the same rule now.

Four fixes, the audible one being a note held by the sustain pedal clicking when you played it again. Its envelope restarted from silence while the note was still ringing at its sustain level.

[The full notes for 1.5.0](https://github.com/benjamindehli/overtonium/releases/tag/v1.5.0)

## 1.4.0, 30 August 2026

Three more factory presets, nineteen in all. **Legato** arrived at the head of the polyphony list, where pressing a key while another is held moves the note rather than starting it again. **One voice per key** became a switch, so a bell struck over and over can be made to build up rather than take over its own voice. Right-clicking a mute or a solo clears every one of them across the mixer, which beats hunting for a stray solo among sixty-six buttons four pixels apart.

**Before upgrading.** Both lists gained an entry at the top, so polyphony indices shifted by one and a session saved on 8 voices opens on 6. The factory presets are renumbered for the same reason.

[The full notes for 1.4.0](https://github.com/benjamindehli/overtonium/releases/tag/v1.4.0)

## 1.3.0, 28 August 2026

**The per-channel outputs are gone again**, and the release is mostly that removal. Logic offers a fixed menu of multi-output arrangements and cannot complete the larger ones, which arrangements work varies with the version and the processor underneath it, and the one shape that would actually suit a 32-partial mixer, 32 mono outputs, was never on the menu at all. A convenience that half worked in the one host it had been tried in is not part of the instrument.

Also fixed: parameters hold the values they are given. Two conversions between a knob position and a value were done in single precision, and 66 defaults fell where that round trip loses a bit.

[The full notes for 1.3.0](https://github.com/benjamindehli/overtonium/releases/tag/v1.3.0)

## 1.2.0, 26 August 2026

An output per channel, 33 of them alongside the stereo mix, which 1.3.0 then removed. MPE slide gained a destination, off, brightness or tuning. The two LINK tilts became one **taper from wherever you grab**, which can peak under your hand in the middle of the series rather than only at an end. The level readout counts down to -99.9 dB rather than giving up at -66, which was two thirds of the fader's usable bottom end. Every control got a name a screen reader can read, all 635 of them.

**Before upgrading.** Automation written against the old level fader plays back at different levels, because the fader's travel is spaced by decibels now rather than by gain.

[The full notes for 1.2.0](https://github.com/benjamindehli/overtonium/releases/tag/v1.2.0)

## 1.1.0, 24 August 2026

**Folding the mixer down.** Click a section heading in the caption gutter and that group of rows folds away across every channel at once, taking the window's height with it. All five groups is 480 pixels shorter.

The factory presets appeared in Logic's own preset menu, the panel gained its grain, its grooves and its bloom, and the [update check](https://benjamindehli.github.io/overtonium/install/) arrived, off unless you ask for it. Nothing about the sound changed, and a patch saved in 1.0.0 loads identically.

[The full notes for 1.1.0](https://github.com/benjamindehli/overtonium/releases/tag/v1.1.0)

## 1.0.0, 23 August 2026

The first release. 32 sine partials and a noise channel, each with its own fader, tuning, envelope, tremolo, pitch modulation and pan. [TUNE](https://benjamindehli.github.io/overtonium/tuning/), six keyboard temperaments, inharmonic stretch, keyboard tracking, drift and wobble. MPE alongside ordinary MIDI, a tape echo, a feedback delay network reverb, a lo-fi converter, channel ganging, and 16 factory presets.

[The full notes for 1.0.0](https://github.com/benjamindehli/overtonium/releases/tag/v1.0.0)

## Being told about a new one

The plugin can check for itself, and does not unless you ask it to. The [install page](https://benjamindehli.github.io/overtonium/install/) says what that costs, which is one request for a static file and nothing about you. Every release is also on the [releases page](https://github.com/benjamindehli/overtonium/releases), which GitHub publishes as [a feed](https://github.com/benjamindehli/overtonium/releases.atom) for anyone who would rather subscribe than be asked.
