Music Source Separation

The song is split into four separate MP3 files: vocals, drums, bass, and all other instruments (other).

You can upload an audio file up to 25MB in size.

For an average 3-minute song, processing on CPU can take about 3-5 minutes; for longer tracks, the time increases.

We support MP3, WAV, M4A, and FLAC formats.

The hybrid transformer model used is one of today's most advanced open-source separation technologies; however, in passages where vocals and instrumental overlap very heavily, slight leakage (a small amount of one part mixing into another) may be heard.

No, it is not recommended. The tool only performs technical separation; it does not transfer copyright. You should only process audio that you own or have legal rights to use, and you must not republish copyrighted content without permission.

Yes, the tool is currently offered for free.

No, the tool processes one file at a time. For multiple songs, you need to repeat the process sequentially.

Our server infrastructure runs on CPU; therefore, processing is slower than professional studio software with a GPU, but it allows it to be offered free to everyone without additional hardware costs.