Text-to-Speech on AISHELL-3 (CER and Acoustic Quality)
10.19CERProposed
Evaluation Results
| Method | Links | |||||
|---|---|---|---|---|---|---|
| ProposedMethodology=Dynamic prosody prediction2026.06 | 10.19 | 82.58 | 66.08 | 92.66 | 5.91 | |
| CosyVoice(50k)Training hours=50k2026.06 | 11.59 | 80.41 | 69.92 | 90.59 | 6.66 | |
| CoTPrompting=Chain-of-Thought2026.06 | 11.61 | 80.61 | 69.9 | 90.51 | 6.52 |