Act 01 · Quick start
In tune in four moves
belt-in is the
live-vocals build (slot synth on the hardware input). belt is
the same engine as a chain / Master FX insert for anything already in your
set — see Two builds.Act 02 · Correction
From invisible to robot
Belt tracks your pitch ~172 times a second, decides the note you mean in the key and scale you set, and glides you onto it. Four knobs shape how honest that correction is:
Retune is the speed of the pull. At 0 the correction is instant — the T-Pain / Cher effect, pitch snapping between notes with no glide. Around 25–40 it reads as a great vocal day; high values only nudge long notes.
Amount is how far toward the target you're pulled. 100 lands on the note; 50 splits the difference; 0 is bypass — handy with harmonies: keep the lead natural, let the backup singers be perfect.
Flex (Shift+knob 2) protects expression: singing between notes — slides, scoops, blue notes — gets left alone, and correction only engages as you commit to a note. Auto-Tune calls this Flex-Tune.
Humanize (Shift+knob 1) splits your pitch into a slow line and the fast wobble on top. Correction moves the slow line; humanize keeps your vibrato riding it instead of ironing it flat. It also loosens the harmony voices — see below.
Act 03 · Harmonies
Backing singers, in key
Four harmony voices ride your corrected pitch at intervals measured in scale degrees — a "+3rd" is a major or minor third as the key demands, the way real backing singers hear it. Each voice gets its own small detune, a slow drift, and a place in the stereo field, so four voices read as four people, not one robot choir.
Pads 1–4 on the bottom row toggle the voices. Shift+pad steps a voice through the intervals:
| Interval | On A in A major | Use it for |
|---|---|---|
| +3rd | C♯ above | the classic close harmony — default for voice 1 |
| +4th / +5th | D / E above | open, modal color; power-chord vocals |
| +6th | F♯ above | sweet, gospel-leaning top line |
| +Oct / −Oct | A above / below | shine or weight; −Oct fakes a bass singer |
| −3rd → −6th | below your note | harmonies under the lead — thicker, darker |
| Unison | your note | with detune + spread = extra body |
Harm Level (knob 5) is the shared voice volume; Spread (Shift+knob 3) fans them across the stereo field; Humanize loosens their tuning and timing drift. Harmonies duck automatically when you stop singing pitched material — backing singers don't harmonize your consonants.
Scales
Chromatic · Major · Minor · Harmonic Minor · Dorian · Mixolydian · Major Pentatonic · Minor Pentatonic · Blues. Chromatic corrects to the nearest semitone (any key, maximum robot); pentatonics are very forgiving live — there are no wrong notes nearby.
Doubler
Knob 6 brings in two extra copies of your corrected voice, detuned a few cents apart and panned hard left and right — the "stacked takes" thickness on every chorus you've ever loved. It costs nothing to leave the harmonies off and just double.
Act 04 · The surface
Every pad, mapped
Tap any pad or step below to see what it does. Rows are counted from the top, the way you actually look at the device.
Ableton Move — pad grid & step row
Knobs 1–8
tuner sets it too
Major default
100 gentle
0–100
0–100
0–100
deep ↔ bright
lead blend
Hold SHIFT — Humanize · Flex · Spread · Monitor
Act 05 · Two builds
Belt In & Belt
| Module | Type | Use it when |
|---|---|---|
| belt-in | instrument slot (audio-in) | Live vocals. Reads the Move's mic / line-in / USB-C directly — load it in a slot and sing. |
| belt | chain / Master FX insert | Processing something already in the set — a sampled vocal, a synth line that wants hard-tune glitch, a whole bus through the doubler. |
Act 06 · Feedback guard
The mic and the speaker
Singing into the Move's mic while its own speaker plays you back is a feedback loop waiting to happen. When Belt In sees speaker on + no line-out cable, it mutes its output and tells you why. Plug in headphones or a line out and it un-mutes by itself.
The Monitor pad (row 4, column 6) is the manual override: green = output on, red = muted. Press it to force monitoring on when you know the room can take it — the guard won't fight you until the jack state changes again.
Act 07 · Help & accessibility
Help on the device, help for everyone
On-device help
Both modules ship a full manual in the Move's own Help viewer — overview, quick start, every knob and pad, feedback-guard notes and tips. No phone required on stage: schwung's Help menu → Belt / Belt In.
Screen reader
With schwung's screen reader enabled, Belt speaks everything it does: knob changes announce full parameter names and values ("Retune speed, 25 percent" — never abbreviations), harmony pads announce the voice and interval ("Voice 1, third up"), the HARD pad, key changes from the tuner strip, and feedback-guard events ("Feedback risk. Output muted…") are all spoken as they happen.
This site
This manual is a single self-contained page built to be readable by everyone: semantic landmarks and headings for screen readers, a skip-to-content link, real buttons with visible keyboard focus, the interactive pad-map readout announced via a live region, reduced motion respected (the marquee stops, reveals don't animate), and color never used as the only signal — every pad and step also says what it is in text.
Found something that doesn't work with your setup? Open an issue on the repo.
Act 08 · Install
Two minutes to first note
Requirements: an Ableton Move with schwung installed.
Via Schwung Manager: open move.local:7700, then
search the Module Store for Belt or Belt In. For a custom install, use:
timncox/schwung-belt-voice — live vocals Voice
timncox/schwung-belt-fx — chain / Master FX
Or by tarball: grab
belt-in-module.tar.gz / belt-module.tar.gz
from their dedicated FX or Voice release page and upload them in
Schwung Manager.
Updating: after an update, fully exit and relaunch the module so the Move loads the new files. Check the version under Help → Modules.
Act 09 · Specs
Under the hood
One pitch analysis feeds every voice: a YIN detector tracks you ~172× a second, a train of pitch marks follows your voice, and all seven output voices — corrected lead, four harmonies, two doubler streams — are PSOLA grain streams reading the same marks. That's why the whole choir costs barely more than one voice. Formant is grain-content resampling, independent of the pitch ratio.