Text-to-Speech on ZTTS1-Eval In-The-Wild (th)
3.84WEREleven Labs V3
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| Eleven Labs V3supports zero-shot voice cloning=false2026.06 | 3.84 | 3.48 | 33.9 | |
| VoxCPM 22026.06 | 4.74 | 2.39 | 78.6 | |
| ZONOS2 8B Quality ModeQuality Mode=true2026.06 | 5.93 | 2.74 | 68.3 | |
| Qwen 3 TTS 1.7B2026.06 | 9.61 | 2.93 | 73.1 | |
| ZONOS2 8BQuality Mode=false2026.06 | 12.87 | 2.43 | 68.4 | |
| Fish S2 Pro2026.06 | 74.12 | 2.75 | 75.5 | |
| Ground truth2026.06 | 86.41 | 2.4 | 82.5 |