Singing Voice Conversion on SVC Evaluation Dataset
15.1PERUniSinger_SVC
Evaluation results
| Method | Links | |||||||
|---|---|---|---|---|---|---|---|---|
| UniSinger_SVCvariant=pure SVC2026.06 | 15.1 | 71.2 | 77.1 | 3.912 | 3.758 | 3.825 | 3.764 | |
| So-VITS-SVC 3architecture=VITS-based2026.06 | 15.4 | 70 | 74.3 | 3.91 | 3.733 | 3.721 | 3.522 | |
| UniSinger_SVC_BGMvariant=SVC with accompaniment co-generation2026.06 | 16.7 | 68.7 | 65.5 | 3.785 | 3.612 | 3.771 | 3.891 | |
| HQ-SVCmode=zero-shot2026.06 | 18.7 | 62.7 | 80.1 | 3.787 | 3.678 | 4.015 | 3.412 | |
| NeuCoSVCframework=neural concatenative2026.06 | 24.3 | 57.3 | 61.2 | 3.824 | 3.765 | 3.899 | 3.237 |