Sound design / AI preset prompt

Describe a sound.
Try it in KIT.

Give your AI assistant the preset guide below, then describe the sound you want. It can write a voice preset for you to load, listen to and make your own.

Voice file format 3 · Updated 5 October 2026

01 / Describe
Start with one sound.
Copy the prompt into your AI chat and add your request. Try “a short, dry snare with a low body and a soft attack.”
02 / Load
Bring it into KIT.
Ask for a .trumvoice file. Use a spare slot or save your project first. In that voice’s preset browser, choose BROWSE PRESETS..., select the file and confirm the file dialog. The preset loads immediately.
03 / Listen
Shape the next version.
Press HIT and listen at a comfortable level. Tell the AI what needs changing: shorter, darker, less ring, more body. Or turn the controls yourself.

Take it to your AI chat

The preset prompt

Download .txt

The guide covers the sound controls, allowed values and file format. Your AI proposes a starting point; your ears decide what works.

Read or select the full prompt

TRUMSYNT-KIT preset-writing prompt

TRUMSYNT-KIT — AI preset-writing prompt
Revision: 2026-10-05. Compatibility: TRUMSYNT-VOICE-FILE 3, 44 PATCH values.

Help me design one sound for TRUMSYNT-KIT, an eight-voice drum synthesizer.
Use the preset interface below to program the existing instrument. Each slot
has the same sound controls. A .trumvoice file describes one sound, not a kit,
pattern or audio recording. Do not substitute TRUMSYNT-16's .ts16 format.

MY SOUND REQUEST
I will describe the sound after this prompt. If I have not done so, ask what
sound I want. Otherwise, make one useful starting preset without a long
questionnaire. Ask one brief question only if an essential detail is missing.
Respond in my language; use a short ASCII preset name and filename.

WORKING TOGETHER
- Design one sound at a time. Describe its intended character, not a listening
  result. Do not claim you heard, tested or matched it unless you actually did.
- Supply a complete .trumvoice file attachment when your tools support it.
  Otherwise put the exact file contents in one plain-text code block and tell
  me to save it as UTF-8 plain text with the .trumvoice extension, not .txt.
  Never put Markdown fences or commentary inside the saved file.
- Briefly explain the main sound choices and name two controls I can adjust
  after listening. Wait for my feedback before making another version.
- If I provide a preset, keep all values I did not ask to change. Return a new
  filename for each revision; do not overwrite my original without asking.
- If I name a drum machine or recording, treat it as inspiration unless I
  provide a specific reference. Do not promise an authentic reproduction.
- Choose a modest VOL for the first audition. GAIN and drive affect the sound;
  use VOL for the listening level. Level cannot be established from numbers
  alone. Ask me to listen at a comfortable monitor level.

HOW TO THINK ABOUT THE SOUND CONTROLS
VCO 1 and VCO 2 supply the pitched body. Each has frequency, waveform, level,
pitch-envelope depth and pitch-envelope decay. The two pitch decays do NOT
give the oscillators independent amplitude envelopes. Positive pitch depth
starts above the base pitch and settles toward it; negative starts below it
and rises toward it. The pitch-envelope shape is shared by both oscillators.

FM sends VCO 1 into VCO 2's frequency modulation. VCO 2 must contribute to the
mix to hear its FM result. VCO 1's mixer level is separate from its role as
the modulator. Hard sync couples the oscillators' cycles. VCO 1 also offers
MTL, a metallic source; VCO 2 does not. These are useful timbre controls, not
an arbitrary waveform editor or a bank of independently shaped resonances.

NOISE has level, dark-to-bright color and its own decay. It can bypass the
voice filter, but still passes through the main amplitude envelope.
CLICK is a short noise-based transient. CLICK TUNE changes its brightness,
not the pitch of another sine oscillator. CLICK has its own decay. With CLICK
VCF BYPASS on, its own envelope can continue outside the main VCA envelope;
with bypass off, it passes through the voice filter and main VCA. RETRIG
repeats the click transient, not the entire voice. Pulse spacing and fall-off
are fixed: do not attempt to edit the compatibility values in fields 42–43.

The voice filter offers LP or HP, cutoff, resonance, envelope depth and
envelope decay. Positive filter-envelope depth initially raises cutoff;
negative depth initially lowers it. NOISE MOD adds movement to cutoff even
when audible NOISE LEVEL is zero. High resonance can dominate the sound.

VCA DECAY controls the body amplitude decay. Start here when a sound rings
too long: changing oscillator pitch decay alone will not shorten its body.
VCA EG changes the onset from fast toward slower. UNTZ shapes both the
amplitude and filter-envelope contours: negative values make them snappier,
positive values rounder and fuller; zero leaves the contours unchanged.
PRE DRIVE and POST DRIVE color the sound before filtering and after the main
amplitude stage respectively; they are not neutral volume controls.

Each voice has one LFO with one destination. Leave it OFF for a simple first
patch unless the request needs movement. SYNC makes RATE choose note values;
with SYNC off RATE is in Hz. RETRIG restarts the LFO on each hit. For a synced
LFO, set field 39 to 1 and use the field 35 note-value mapping below.
LFO TUNE moves both VCOs; performance TUNE also moves CLICK.

Practical starting choices, not recipes for specific machines:
- Kick: a low pitched body, a restrained pitch sweep and a short transient.
- Snare: pitched body plus noise; adjust body length and noise contribution
  separately, remembering that the noise also follows the main VCA.
- Clap: noise tail plus repeated CLICK; compare bypass on/off after listening.
- Hat: noise and/or MTL, filtering and a short or longer amplitude decay.
For a tighter sound, try reducing VCA DECAY before adding more drive or FM.
There is no sample playback, reverb or arbitrary envelope editor in this
preset interface. Do not invent additional parameters.

FILE CONTRACT
Use exactly these five lines, ending with a newline:
TRUMSYNT-VOICE-FILE 3
NAME <short name>
PATCH <exactly 44 space-separated numeric values in the order below>
PERF <TUNE position> <DECAY position> <VEL position>
VOL <slot volume position>

NAME: use 1–14 printable ASCII characters, no newline. Use the same base name
for the filename. Prefer descriptive names such as Soft Knock or Dust Snare;
do not assume my library has a particular free preset number.
Numbers: decimal point, no units, commas, NaN or infinity. Boolean fields are
integer 0 or 1. Choice fields are integer indexes, never waveform names.
PATCH is one physical line. Do not add field labels, comments or extra values.

PATCH FIELD ORDER AND ALLOWED VALUES
01 | VCO 1 pitch decay       | 0.01–2 seconds
02 | VCO 2 pitch decay       | 0.01–2 seconds
03 | VCA decay               | 0.01–2 seconds
04 | VCF decay               | 0.01–10 seconds
05 | VCO 1 frequency         | 1–8000 Hz
06 | VCO 2 frequency         | 1–8000 Hz
07 | VCO 1 level             | 0–1
08 | VCO 2 level             | 0–1
09 | VCO 1 waveform          | 0=SQR, 1=TRI, 2=MTL, 3=SIN
10 | VCO 2 waveform          | 0=SQR, 1=TRI, 3=SIN (do not use 2)
11 | VCO 1 pitch depth       | -60–60 semitones
12 | VCO 2 pitch depth       | -60–60 semitones
13 | Pitch envelope shape    | 0=EXP, 1=LIN, 2=LOG
14 | FM amount               | 0–1
15 | Hard sync               | 0=off, 1=on
16 | Noise level             | 0–1
17 | Noise color             | -1–1; negative darker, zero neutral, positive brighter
18 | Noise decay             | 0.005–2 seconds
19 | Noise VCF bypass        | 0=off, 1=on
20 | Click tune              | 20–8000 Hz
21 | Click decay             | 0.005–0.2 seconds
22 | Click level             | 0–1
23 | Click VCF bypass        | 0=off, 1=on
24 | Filter cutoff           | 20–20000 Hz
25 | Filter resonance        | 0–1
26 | Filter mode             | 0=LP, 1=HP
27 | Filter envelope depth   | -1–1
28 | Stored compatibility    | keep 0.5 for new presets; preserve in supplied presets
29 | Noise filter modulation | 0–1
30 | VCA onset               | 0–1; zero is fastest
31 | UNTZ                    | -1–1; negative snappier, positive rounder, zero neutral
32 | Stored compatibility    | keep 0.84 for new presets; preserve in supplied presets
33 | Pre drive               | 1–10; start at 1
34 | Post drive              | 1–10; start at 1
35 | LFO rate                | 0.01–20 Hz with sync off; with sync on use mapping below
36 | LFO waveform            | 0=TRI, 1=SAW, 2=SQR, 3=SH
37 | LFO destination         | integer 0–29; mapping below
38 | LFO amount              | 0–1
39 | LFO sync                | 0=off, 1=on
40 | LFO retrigger           | 0=off, 1=on
41 | Click retrigger pulses  | integer 1–8, including the first pulse
42 | Stored compatibility    | always 0.01325
43 | Stored compatibility    | always 0.78
44 | Voice gain              | 0–2; start at 0.8

SYNCED LFO RATE — PATCH FIELD 35
With SYNC on (field 39 = 1), field 35 stores cycles per quarter-note beat,
not Hz or a choice index. Use one of these values:
0.0625=4 bars; 0.125=2 bars; 0.25=1 bar; 0.5=1/2;
1=1/4; 2=1/8; 4=1/16; 8=1/32.
Bar labels assume four quarter-note beats per bar. Only straight note values
are available. For example, an eighth-note LFO uses field 35 = 2 and field
39 = 1; keep field 37 set to the intended destination and field 38 above zero
to hear modulation.

LFO DESTINATION INDEXES
0 OFF; 1 CUTOFF; 2 RES; 3 VCF EG; 4 NOISE VCF; 5 FM AMT;
6 VCO1 PITCH; 7 VCO2 PITCH; 8 NOISE; 9 NOISE COL; 10 CLICK LVL;
11 VCO DEC; 12 VCO2 DEC; 13 VCF DEC; 14 VCA DEC; 15 VCA EG;
16 PRE DRV; 17 POST DRV; 18 UNTZ; 19 CLICK TUNE; 20 NOISE DEC;
21 CLICK DEC; 22 RETRIG; 23 VEL; 24 PAN; 25 TUNE; 26 VCO1 LVL;
27 VCO2 LVL; 28 VCO1 PITCH EG; 29 VCO2 PITCH EG.

PERFORMANCE AND VOLUME
PERF has three positions, each 0–1, in the order TUNE, DECAY, VEL.
Use PERF 0.5 0.5 0.8 for a new starting sound. TUNE 0.5 is untransposed;
its endpoints are -12 and +12 semitones. DECAY 0.5 leaves the programmed
decays unchanged; its endpoints are x0.25 and x4. These controls modify the
sound when played; do not bake their effect into PATCH a second time.
VOL is 0–2: 0 is silent, 1 is unity, 2 is +12 dB. It is a knob position, not
a linear gain multiplier. Start at 0.65. Do not compensate for every sound
with high GAIN or drive. VEL, accents, step locks and the DAW mix also affect
what I hear; a preset file does not set the whole kit or the master bus.

COMPLETE STARTING EXAMPLE — Soft Knock.trumvoice
This is an editable starting sound, not the factory Init preset.

TRUMSYNT-VOICE-FILE 3
NAME Soft Knock
PATCH 0.025 0.05 0.045 0.15 70 180 0.6 0 3 3 5 0 0 0 0 0 0 0.05 0 1800 0.01 0.06 1 5000 0 0 0 0.5 0 0 0 0.84 1 1 1 0 0 0 0 0 1 0.01325 0.78 0.8
PERF 0.5 0.5 0.8
VOL 0.65

BEFORE RETURNING A FILE
Count all 44 PATCH values and all three PERF values. Check every field's
position, range and integer/boolean requirements. Keep fields 28 and 32 at
their starting values in new presets and fields 42–43 at their fixed values.
Check that the intended audible sources are enabled and that GAIN and VOL
are not zero. Do not claim a parser or audio check unless it was performed.

HOW I LOAD THE RESULT
Save the .trumvoice file to Documents/TRUMSYNT-KIT/Presets, or another chosen
folder. Use a spare slot or save your project before auditioning: loading
replaces that slot's sound and VOL and clears its synth/performance step locks
in all patterns. In KIT, open the selected voice's preset browser, choose
BROWSE PRESETS..., select the file and confirm the file dialog. The preset
loads immediately; no separate LOAD step is needed. HIT plays the sound.
Save useful adjustments in KIT.
If KIT refuses the file, check the extension, five-line structure, format
version, numeric values and PATCH count before changing any sound settings.
If it loads but sounds wrong, ask what I hear and change only a few relevant
controls in the next version. My listening is the final judgement.

MY REQUEST:

A starting file

Soft Knock

A short pitched body with a small click. Use it to try loading a preset, or attach it to your AI chat and ask for a variation.

Download preset

Saving and auditioning

If your AI returns text, save it as plain text with the .trumvoice extension. Keep only the file contents, without the chat’s code fences.

You can keep your sounds in Documents/TRUMSYNT-KIT/Presets. Loading a voice replaces that slot’s sound and volume and clears its synth and performance step locks in all patterns. Use a spare slot or save your project first.

Voice presets in the manual ↗

Start with one sound.

Bring it into KIT.

Shape the next version.

The preset prompt

Soft Knock

Saving and auditioning