What is a Stem Splitter?
A stem splitter uses AI to separate a finished song into its main parts. This tool creates four WAV stems: Vocals, Drums, Bass, and Music. Music is the remaining instruments and sounds that are not vocals, drums, or bass.
The model runs on your device with WebGPU, a newer browser feature that uses your graphics processor. The song is never sent to a server. Temporary audio stays in private browser storage and is cleaned up when you start another split.
How to Split a Song into Stems
1. Choose an audio file
Open a song up to six minutes long. The tool checks its duration and available private storage before downloading the AI model.
2. Let the AI separate it
Watch model download, graphics health check, separation, and finalizing progress. Everything happens in this browser tab.
3. Preview and download
Solo or mute stems, seek through the streamed preview, then download individual WAV stems or a minus-one version.
Frequently Asked Questions
Does my song get uploaded?
No. The AI model, separation, playback, and downloads all run locally in your browser. Your audio never leaves your device.
Which browsers work with the Stem Splitter?
Use a recent Chrome or Edge browser on a computer. The tool requires WebGPU. iPhone and iPad support is coming soon, and there is no slow CPU or server fallback.
What is the Music stem?
Music is the model's remaining-instruments bucket. It commonly contains guitars, keyboards, strings, synths, and other sounds that are not vocals, drums, or bass.
How long can my song be?
The current limit is six minutes. Duration and storage checks happen before the large AI model download begins.
What can I download?
You can download Vocals, Drums, Bass, and Music as separate WAV files, plus Drumless, Bassless, Instrumental, and Acapella WAV files.
More private browser music tools
Need a simpler two-track vocal split, or want to edit the song itself?