The AI behind the vocal remover
FreeVocalRemover runs Hybrid Transformer Demucs, an open-source music source-separation model from researchers at Meta AI. Instead of simply filtering out the frequencies where voices usually sit — the trick old karaoke "vocal cut" tools used — it was trained on hundreds of multitrack studio recordings and learned what voices and instruments actually sound like. That is why it can remove harmonies and ad-libs too, while keeping the bass, drums and the rest of the arrangement intact.
Free limits and privacy
- 3 free separations per day, no account needed.
- Files up to 6 minutes long and 50 MB in MP3, WAV or M4A format.
- Uploads and results are deleted automatically after 1 hour.
- Only upload audio you own or have permission to use.