Text-to-Speech on ZTTS1 Eval In-The-Wild (id)
5.98WEREleven Labs V3
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| Eleven Labs V3supports zero-shot voice cloning=false2026.06 | 5.98 | 3.34 | 26.2 | |
| VoxCPM 22026.06 | 6.43 | 2.3 | 76.9 | |
| ZONOS2 8B Quality ModeQuality Mode=true2026.06 | 6.48 | 2.66 | 67.6 | |
| Fish S2 Pro2026.06 | 11.02 | 2.72 | 71.4 | |
| ZONOS2 8BQuality Mode=false2026.06 | 11.84 | 2.25 | 71.9 | |
| Qwen 3 TTS 1.7B2026.06 | 12.46 | 2.94 | 67.4 | |
| Ground truth2026.06 | 17.23 | 2.14 | 80 |