Users who need cleaner vocals before practice, dubbing, or remix work
Vocal and accompaniment stems for practice, sampling, transcription, or cover workflows
Input
Full songs, cover material, or vocal-with-backing files; separated first with RoFormer stems
Audio formats
Output
Vocal and accompaniment stems for practice, sampling, transcription, or cover workflows
Best for
Users who need cleaner vocals before practice, dubbing, or remix work
Vocal/accompaniment separation is now a fixed RoFormer separator chain. It creates vocals with harmony, accompaniment without harmony, original vocals, and accompaniment with harmony without invoking YourMT3.
Primary vocal/instrumental pass. TelkNet uses an audio-separator RoFormer ensemble optimized for vocal/RVC-style separation, not a transcription model.
Second-pass lead/back vocal separation. It runs only on the extracted vocal stem so backing vocals can be preserved as harmony-aware accompaniment.
Runtime wrapper for UVR/MVSep-style separator checkpoints. It provides the model loading, WAV stem export, and FFmpeg-backed audio format handling used by this tool.
arXiv / ISMIR late-breaking demo - 2023
Mel-RoFormer replaces heuristic non-overlapping bands with overlapping mel-band subbands and reports better separation than BS-RoFormer for vocals, drums, and other stems on MUSDB18HQ.
SDX23 music separation track - 2023
BS-RoFormer uses band-split spectrogram features plus RoPE Transformers; the SDX23 system ranked first in the music source separation track and a smaller version reported 9.80 dB average SDR on MUSDB18HQ.
TISMIR - 2024
The SDX23 music demixing paper documents robust MSS benchmarks, listening tests, and that the best system improved over the prior challenge by more than 1.6 dB SDR on MDXDB21.