Skip to content
Audio to notation
Front-end demo

Transcribe your song.

Give Timbre a recording and it writes out the notes. Piano, guitar, bass and strings each get their own stave.

Any source

Paste a link, drop a file, or play into the mic. Timbre converts it to WAV before anything else touches it.

Per instrument

Piano gets a dedicated onset model. Everything else goes to a pitch tracker. Each one lands on its own stave.

Yours to edit

Every run gives you MusicXML, MIDI and PDF. Open the MusicXML in MuseScore and fix what you want.

Inside
One honest caveat

Nobody has solved polyphonic transcription yet, and Timbre has not either. What we have here is a working draft, not a finished engraving. Rhythms will need fixing and the voicing will be wrong in places. That is why every run hands you the MusicXML.

About this deployment

What you are looking at is the front end only. The transcription pipeline needs PyTorch, ffmpeg and LilyPond on an always-warm machine, so this build replays a finished job instead of running one.

The score, MIDI and PDF you can download are not mocked. They came out of the real pipeline and from a real recording, and were copied into this build.

Get in touchfor the full build, or to run it on your own material