Getting Started
Drag and drop audio, click to upload, or record below
Any format the browser can decode works — the page converts it to wav before it is sent, so the studio never needs a codec it does not have. Recording captures raw PCM in the browser, so a take is exact, not re-encoded.
Add voices
Result
What was measured
| nothing converted yet |
Discover
Train a RVC Voice
Build a clean training dataset from microphone takes or uploaded audio, then open the RVC trainer.
Clone Voice
Clone someone's voice from a recording in voices/, or design a new one from the catalogue.
Create Demo with Your Voice
Take this voice into the studio and sing a song with it.
Isolate Vocals
Keep the centre of a stereo mix — where a lead vocal lives — and attenuate what is panned wide. Needs a stereo file.
Create Vocal Layers
Stack detuned, delayed copies of the take into a choir of itself.
Clean the take
Measure the noise floor bin by bin across the whole file and subtract it, keeping a floor so the gaps do not chirp. The report says how far the floor fell and how much of the voice stayed.
Filter rack
Nine stages, listed by the engine rather than by this page. Tick the ones you want and they run in the order shown; each reports what it measured, not what it was asked for.
Lyric filters
Before anything is sung: a blocked word is replaced, a word with no vowel is dropped, a line already in the song goes, and a line too long for a bar is shortened at a word boundary. The receipt names every change.