⌘K

AI Stem Separation

Split any song into clean vocals, drums, bass and instruments in your browser. Free, no install — runs on Demucs AI.

  • ~1-3 min processing
  • Up to 64 MB
  • MP3, WAV, FLAC, M4A, OGG
  • Files auto-deleted after processing

Separation Configuration

Select the number of stems to separate and choose the track you want to isolate.

Upload Your Music

Upload your audio file to start the AI stem separation process

Drag and drop your audio file here

Supports MP3, WAV, FLAC, and other audio formats

What is AI stem separation?

Stem separation uses a deep neural network (Demucs) to take a finished mix and split it back into its source tracks - vocals, drums, bass and other instruments - as separate audio files. The model was trained on thousands of multi-track recordings and runs on our backend, returning each isolated stem as a clean WAV/MP3 you can download or remix.

Use it to build karaoke versions, extract acapellas for sampling, lift drum loops, transcribe a bassline, or feed individual stems into a DAW for remixing. The quality is studio-grade for vocals and drums; bass and "other" depend on the source mix density.

What can you do with separated stems?

Three popular workflows have their own dedicated entry point with the right preset selected.

Make a karaoke track

Remove the lead vocal and keep the full instrumental backing — perfect for singing along.

Open Karaoke Maker

Extract an acapella

Isolate just the vocals from any song — for sampling, remixing or vocal practice.

Open Acapella Extractor

Get a drumless track

Strip out the drums to practice along on your kit, or to replace them with your own.

Open Drumless Tracks

Continue your workflow

Once your stems are separated, take them straight into another tool.

Audio Editor

Cut, loop, change BPM and pitch on each stem in a multi-track timeline.

Looper

Layer your stems live and build looping arrangements on top of them.

Backing Tracks

Generate chord, bass and drum jam tracks to pair with your isolated stems.

Tuner

Tune your instrument before recording over an instrumental stem.

Frequently asked questions

How long does separation take?
Most 3-4 minute songs finish in 1-3 minutes. Time depends on the file length and current backend load — long tracks (10+ minutes) can take up to 5 minutes.
What audio quality do I get?
Stems are separated and streamed back as WAV (lossless) and you can also export a remixed MP3 at 192 kbps. Best quality is on vocals and drums; bass and 'other' depend on how dense and well-mixed the source is.
Is my file uploaded somewhere permanent?
Files are processed on our backend and auto-deleted after the job completes. We don't store your audio long-term and don't use it to train any model.
What's the maximum file size?
Up to 64 MB per upload. For most songs at 320 kbps MP3 that's about 25 minutes of audio.
Should I pick 2, 4 or 6 stems?
Pick 2 if you only need one specific instrument isolated (e.g. just vocals removed for karaoke) — it's the fastest. Pick 4 for the standard vocals/drums/bass/other split. Pick 6 if you need separated piano and guitar in addition.
Does it work on every genre?
It works best on modern produced music (pop, rock, hip-hop, electronic). Highly orchestral, classical, choir-heavy or live-recorded mono tracks can produce weaker separation because the model is trained on contemporary multi-track stems.