Multi-speaker Voice Clone on JoyVoice MSMT-eval en (test)
3.36WER (2spk)JoyVoice
Evaluation Results
| Method | Links | ||||||
|---|---|---|---|---|---|---|---|
| JoyVoiceZero-shot=true, Decoding Source=official2025.12 | 3.36 | 3.61 | 6.01 | 9.57 | 3.93 | 12.18 | |
| VibeVoice-7BZero-shot=true, Decoding Source=open-source model2025.12 | 4.19 | 6.58 | 6.66 | 10.6 | 6.42 | 13.81 | |
| VibeVoice-1.5BZero-shot=true, Decoding Source=open-source model2025.12 | 4.26 | 7.73 | 8.56 | 12.22 | 8.89 | 14.11 | |
| MooncastZero-shot=true, Decoding Source=open-source model2025.12 | 5.32 | 17.38 | — | — | — | — | |
| MOSS-TTSDZero-shot=true, Decoding Source=open-source model2025.12 | 5.56 | 10.04 | — | — | — | — |