Emotional Dialect Synthesis on Emotional Dialect Synthesis dataset
2.75TianjinFully E-Vector
Evaluation results
| Method | Links | |||||||||
|---|---|---|---|---|---|---|---|---|---|---|
| Fully E-VectorMerging Strategy=fully merging2025.12 | 2.75 | 2.09 | 2.99 | 2.97 | 3.06 | 2.63 | 2.5 | 3.08 | 2.76 | |
| HE-VectorMerging Strategy=hierarchical merging2025.12 | 2.68 | 1.73 | 3.07 | 2.61 | 3.28 | 2.8 | 3.22 | 3.26 | 2.83 | |
| Dual-stagePipeline=dialect-enhanced F5-TTS and emotion-enhanced F5-TTS2025.12 | 2.31 | 2.04 | 2.61 | 2.86 | 2.83 | 2.88 | 2.28 | 2.7 | 2.56 | |
| CosyVoice2Type=instruction-based multi-style2025.12 | 1.74 | 2.6 | 2.13 | 2.74 | 1.59 | 1.35 | 1.65 | 1.18 | 1.87 |