Voice Style Conversion on StyleStream (test)
3.8WER (%)Ground Truth
Evaluation Results
| Method | Links | ||||||||
|---|---|---|---|---|---|---|---|---|---|
| Ground Truth2026.02 | 3.8 | — | — | — | 3.71 | — | — | — | |
| StyleStreamZero-shot protocol=true, Processing mode=offline2026.02 | 9.2 | 0.852 | 0.64 | 0.827 | 3.42 | 4.32 | 4.42 | 4.36 | |
| CosyVoice 2.0Zero-shot protocol=true2026.02 | 9.5 | 0.794 | 0.45 | 0.655 | 3.47 | 2.26 | 2.58 | 3.25 | |
| StyleStreamZero-shot protocol=true, Processing mode=streaming2026.02 | 15.3 | 0.855 | 0.635 | 0.803 | 3.34 | 4.28 | 4.37 | 4.29 | |
| FACodecZero-shot protocol=true2026.02 | 15.5 | 0.763 | 0.408 | 0.668 | 2.55 | 2.02 | 2.76 | 2.98 | |
| VevoZero-shot protocol=true2026.02 | 17.5 | 0.818 | 0.596 | 0.712 | 3.38 | 3.93 | 3.49 | 3.76 | |
| Vevo 1.5Zero-shot protocol=true2026.02 | 20.5 | 0.798 | 0.554 | 0.683 | 3.1 | 3.15 | 3.15 | 3.34 | |
| SeedVC v2Zero-shot protocol=true2026.02 | 21.7 | 0.766 | 0.549 | 0.688 | 2.65 | 3.34 | 3.21 | 3.11 |