Suu is a talking filter: any sound through a simulated human vocal tract. Shape the mouth on a vowel pad, set it moving, type words for it to sing, or sequence vowels and notes. In the browser, and as a VST3 and CLAP plug-in.
See it live: https://janne-s.github.io/Suu/ · Plug-in: Releases
The sound enters the airway as airflow and leaves through the lips, so the mouth's shape filters it as it would a voice. The airway, articulators and vocal folds are a physical simulation. Tilt, the sub layer, the resonator, glide and Space are production extras. Ranges go past human anatomy on purpose.
- Vowel pad: drag the point to choose the vowel. The pad's corners are
i, u, æ and ɑ.
- Left to right: from front vowels (i, æ) to back vowels (u, ɑ).
- Top to bottom: from close vowels (i, u) to open vowels (æ, ɑ).
- The border outside the corners goes past those vowels, further than a human mouth can.
- Horn: the tract from the side, glottis to lips, as it sounds now.
- Drag to draw your own, build it from a Wave (cycles and depth), or take one period of the sound with the camera. ↺ restores the original.
- Morph blends the horn in.
- Shift slides it along the tract. A wave wraps, so it can travel through.
- Open widens it.
- Text: type words, and choose how they are said.
- Step: each note says the next part, or the text's own Clock does (a rate, or ♪ for a note value).
- Unit: a syllable or a word at a time.
- Say: Hold keeps the vowel while the note lasts, Shot says the part through.
- Len: on the Clock, how much of each tick Hold keeps the vowel. At 100% the parts join.
- Start: Free, or Seq to start over with the sequence.
- Repeat: Loop, or 1× once a pass.
- Speed: how fast it is spoken, from ×0.25 to ×4.
- Gate: how much of each part is heard. Fit: the text bounds the note, so each note lasts only as long as its syllable or word; on the Clock the gaps between parts fall silent.
- Round: fades each part in and out like a grain. At 0% the sound stays on.
- Motion: moves Vowel X/Y, Round, Nasal, Inject, Damping, Morph, Shift
and Open. Each has:
- Depth, and Rate or ♪.
- A shape: sine, triangle, saw, square, random, envelope or ADSR.
- Smooth, from sharp jumps to gliding ones.
Sixteen steps, each with a note, vowel, accent ◆ and length. Drag a step up or down for its note, click to turn it on, Shift-click to tie, click the vowel to change it. Gate scales every step's length alike.
- Open: what opens the sound while the sequencer plays.
- Seq: the steps alone.
- × Keys: the steps while a key is held; the first key starts the pattern. Keys transpose it.
- + Keys: keys also between the steps.
- × In: the steps while the input is above the gate's threshold.
- + In: the input also between the steps.
- Vow: how far the steps' vowels move the mouth from the pad's vowel.
- Osc and In — The oscillator (supersaw, saw, pulse or off) and an input (a file or the mic; in the plug-in, the track) sound together, each at its level. Keys, Hold and the sequencer gate both. ▾ next to the oscillators sets the pulse's Width (50% is a square) and how it swings: Depth, and Rate or ♪.
- Body — Size of the mouth (0.4–4×), Breath (lung pressure for the consonants and folds) and Follow (how far it follows the sound). The quality button steps through HQ, Lean (two to three times lighter) and Lo (lighter still; fast mouth moves are a little smoother).
- Air — The airway's turbulence, each source on its own: Aspir (breath at the folds, h), Fric (the narrowest point: f, sh) and Sibil (the teeth: s), 0–1000%. Whistle tames the ringing of a mouth closed at both the tongue and the lips. Move is the sound of the mouth changing shape, the air its moving walls push: 0% silences it, up to 800%.
- Source — Osc and Input levels, Folds (the model's own vocal folds at the played note; 100% is about as loud as Osc at 0 dB), Reson (a resonator tuned to the played note, so an input plays in pitch) and Glide (pitch and vowel; ♪ syncs it; ↕ ↑ ↓ both ways, only up or only down). With the pitch from the input, Steady takes a new note only once it has lasted, so a slide between notes is not heard as the notes on the way.
- Tone — Drive into the mouth, Tilt (source brightness), and Sub (the dry low end).
- Stereo — A second mouth on the right, a little larger (Spread) and with its vowel moved (Vowel); Width sets how far apart they sound. Phase runs its Motion ahead in the cycle, Canon has it say the text that many parts behind (up to the text's own count), and Mirror turns its horn round. It doubles the CPU, so it is off until switched on.
- Space — Delay and reverb, web only. Off is what the plug-in does.
Keys A W S E D F T G Y H U J K play, Z X change octave, Space plays the sequencer, Esc releases all notes. Settings are kept in the browser. The Preset menu has A/B to compare, Init, and Save and Open for preset files.
macOS 11 or later, Apple silicon. Two plug-ins: Suu (an effect: the track through the mouth, MIDI from another track) and Suu Synth (an instrument: the oscillator through the mouth, a sidechain as input).
In the plug-in the MPE menu chooses what a key's pressure and slide move (any Motion target, Breath or Level, with a depth) and sets the pitch bend range.
Unzip the release and copy the .vst3 files to ~/Library/Audio/Plug-Ins/VST3/
(and the .clap files to ~/Library/Audio/Plug-Ins/CLAP/). The build is not
notarized, so clear the quarantine once (the second line for CLAP):
xattr -dr com.apple.quarantine ~/Library/Audio/Plug-Ins/VST3/Suu*.vst3
xattr -dr com.apple.quarantine ~/Library/Audio/Plug-Ins/CLAP/Suu*.clapRust with the wasm32-unknown-unknown target, and Bun (only for the local
server and the browser checks; the page itself has no build step).
bun install
bun run web # http://127.0.0.1:5318/
bun run build:plugin # VST3 and CLAP in plugin/target/bundled/bun run test # the engine, the core and the plug-in (cargo test)
bun run e2e # headless Chrome, with bun run web running| Folder | What |
|---|---|
crates/chimera/ |
CHIMERA's vocal-tract engine |
crates/suu/ |
Suu's core: the driven airway, oscillator, Motion, text, sequencer |
crates/suu-wasm/ |
the core's interface for the web page |
plugin/ |
the VST3/CLAP effect and instrument |
web/ |
the page, which is also the plug-in's window |
data/ |
pronouncing dictionary and fonts, with their licenses |
tools/ |
web build, local server, browser checks |
- English text only.
- One note at a time.
- A stereo input is mixed to mono. The output is stereo only with Stereo on.
- The page is tested in Chrome and Safari.
- The plug-in is VST3 and CLAP for macOS on Apple silicon, tested in Ableton Live.
Built on CHIMERA by Jameson
Hodge, not affiliated. Suu keeps CHIMERA's engine (crates/chimera/) and its
history, and adds the driven airway, the effect core, the web page and the
plug-in. CHIMERA's own voice app is not part of Suu; it remains in the original
repository. See AGENTS.md for the engine's rules.
Copyright © 2026 Janne Särkelä. Suu is free software under the GNU General Public License v3.0 or later: you may use, change and share it, also commercially, as long as what you share stays under the same license with its source.
CHIMERA's engine (crates/chimera/) is © 2026 Jameson Hodge under the
MIT License. The CMU Pronouncing Dictionary and the
AVHershey fonts keep their own licenses (data/, data/fonts/).
Support development through GitHub Sponsors or Ko-fi. See Sanara Creations for related work.