Uncompose

An open, local-first toolkit for transforming audio, preserving its provenance, and evaluating the results.

Three tools

Uncompose
Transform recorded music into stems.
Uncompose Project
Record the assets, relationships, and processes in an audio project.
Uncompose Compare
Evaluate audio assets through controlled listening and structured observations.

One command namespace

uncompose separate song.wav
uncompose project init
uncompose project import song.stems/job.json
uncompose compare vocals-a.wav vocals-b.wav

Each tool installs and runs on its own; installed together they answer to uncompose.

Local-first, and honest about it

Your audio never leaves your machine. There are no accounts, no telemetry, and no network I/O beyond the separation models you ask Uncompose to download. Your files, your models, your machine.

This page keeps the same promise: static HTML and CSS, no analytics, no third-party requests, nothing to consent to.

You are responsible for having the rights to the audio you process. Separating a recording does not grant you any rights to it.

Get it

Project and Compare are pre-v0.1: their packages are reserved placeholders until the first release. Uncompose itself installs today.