From your rack to anyone's DAW.
A sound signature is a trained model of a real signal chain: one unit or a whole chain of them, hardware, plugins or both. Here's what happens from capture to playback, and how we prove it's accurate.
Capture
Connect your gear to an interface send and return, open the DigMusiK plugin and start the capture assistant.
- The assistant plays a carefully designed reference signal through your gear: sweeps at several levels, level steps, bursts and recoveries that reveal attack and release, drums, tones and music.
- For working knobs, it then asks for short passes (about a minute each) at the settings it picks: "Set Attack to 3, Ratio to 8, press Next". It deliberately covers the extremes, where models are usually weakest.
- Every pass is checked as it arrives: timing, level, clipping, noise and whether the chain was bypassed. Problems are explained in plain English with what to change.
- Plugin chains are captured the same way, and the assistant can turn a plugin's knobs for you.
- Prefer your DAW? Download the reference, render it through your chain and upload the file instead.
What a capture needs
| Capture | Your time | Counts as |
|---|---|---|
| One setting | about 5 minutes | 1 capture |
| Working knobs | about 30 to 60 minutes | 3 captures |
Working-knob captures are included with Studio and Pro. Any audio interface with a spare output and input works; 24-bit at 48 kHz is ideal.
Train
Our engine learns your unit on dedicated GPUs, usually in under half an hour.
- Tone and harmonics: a waveform network learns the unit's frequency response, saturation and how its colour changes with level.
- Dynamics: a second path follows the music's level over milliseconds to seconds, so compressors, limiters and levelling amps keep their attack, release, memory and feel.
- Controls: the knob positions are inputs to the model, so it learns how each control changes the sound, including between the settings you recorded.
- Dial correction: real dials are hard to set exactly. The engine works out where each dial really was from the sound itself and corrects small differences.
How we measure accuracy
Part of every capture is held back and never trained on. The finished signature is scored on that held-back audio, and on settings it didn't train on, at its worst setting rather than its average.
| Rating | Error to signal |
|---|---|
| Excellent | under 1% |
| Great | under 3% |
| Good | under 6% |
| Fair | under 15% |
We also publish the level error in decibels, which shows how closely compression is followed. Only signatures rated Good or better can be published to everyone.
Play
Load signatures in the DigMusiK plugin in any DAW.
- No added latency, and smooth, click-free knob movement
- Rack: chain signatures in any order; Blend: mix any two in any proportion
- Easy mode (Drive, Tone, Mix, Output) on top of the unit's own controls; Advanced adds EQ before and after
- Every value visible and exact: click a knob to see it, Shift-drag for fine control, double-click to type, Alt/Option-click to reset
- AI matching sets up a chain from a reference track
What it can't do (yet)
We'd rather tell you now than in a review.
- Reverb and delay tails longer than about a second are shortened
- Movement from an internal LFO or clock (chorus, flanger, rotary speaker, tape wow) isn't reproduced; the tone is
- Noise that doesn't follow the input (hum, hiss) isn't captured, so signatures are cleaner than the unit
The capture check warns you about each of these before training starts.
Secure by design
Signatures and creators' plugins only run inside DigMusiK, signed in to an account that's allowed to use them. Files are encrypted for the devices you've authorised, never stored unencrypted, and carry an invisible owner mark, so creators' work is protected.