sgmse for vocal source extraction
This website contains supplementary material to the paper:
- An Analysis of Score-based Generative Modelling for Vocal Source Extractiont
| Filename | Clean | Mastered Noisy | HTDemucs | Hybrid Demucs | SGMSE+ |
|---|---|---|---|---|---|
| Arise - Run Run Run | |||||
| Ben Carrigan - We'll Talk About It All Tonight | |||||
| Speak Softly Like Horses | |||||
| Bobby Nobody - Stitch Up | |||||
| The Mountaineering Club - Mallory | |||||
| James Elder & Mark M Thompson - The English Actor | |||||
| Triviul feat The Fiend - Widow | |||||
| Lyndsey Ollard - Catching Up |
References
[1] Bunlong Lay, Jean-Marie Lemercier, Tom Schimansky, Timo Gerkmann "An Analysis of Score-based Generative Modelling for Vocal Source Extraction", preprint