Emotional Text-to-Speech on LSSED and EmoVoice-DB (test)
4.02WERHPRO
Evaluation results
| Method | Links | ||||||
|---|---|---|---|---|---|---|---|
| HPROzero-shot=true2026.06 | 4.02 | 4.171 | 3.65 | 0.339 | 0.672 | 3.73 | |
| CosyVoice3zero-shot=true2026.06 | 4.9 | 4.137 | 3.538 | 0.275 | 0.611 | 3.72 | |
| HD-PPTzero-shot=true2026.06 | 4.92 | 4.068 | 3.547 | 0.323 | 0.646 | 3.75 | |
| CosyVoice2zero-shot=true2026.06 | 5.45 | 4.094 | 3.53 | 0.307 | 0.613 | 3.76 | |
| IndexTTS2zero-shot=true2026.06 | 6.74 | 4.026 | 3.692 | 0.293 | 0.526 | 3.53 | |
| TokenReconmode=reconstruction from ground-truth tokens, zero-shot=true2026.06 | 7.34 | — | — | 0.57 | 0.775 | 3.58 |