splittermixer.domainreviews.org

AI source separation + browser mixer · grab the parts you want, drop the rest, import your own

AI Model · HTDemucs v4 (6-stem)

Click to download HTDemucs v4. The model is then served from your browser's cache on subsequent visits.
Drop a WAV / MP3 / FLAC / OGG here

or click to choose · model must be loaded first

How it works. Drop a song → click Load model if you haven't yet → the page runs HTDemucs v4 in your browser via onnxruntime-web (model from MansfieldPlumbing/Demucs_v4_TRT on Hugging Face). The track is split into drums, bass, other, vocals, guitar, piano, each appearing as its own row in the mixer. Click Mute to silence a part you don't want, Solo to isolate one, drag the volume / pan sliders, or click Discard to remove a track entirely. + Add Track imports any audio file (a different stem, a sample, your own vocal take) as an additional row — synced to the same playback transport. Mix down WAV renders all currently-active tracks (respecting mute/solo/vol/pan) into a single stereo WAV sized to the longest track. Everything runs locally — no audio is ever uploaded.