Text-to-Speech on ZTTS1-Eval In-The-Wild (es)
1.58WEREleven Labs V3
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| Eleven Labs V3supports zero-shot voice cloning=false2026.06 | 1.58 | 3.34 | 16.4 | |
| Qwen 3 TTS 1.7B2026.06 | 1.77 | 2.87 | 65.8 | |
| Gemini 3.1 Flashsupports zero-shot voice cloning=false2026.06 | 1.94 | 3.41 | 20.5 | |
| ZONOS2 8B Quality ModeQuality Mode=true2026.06 | 2.1 | 2.72 | 63.3 | |
| Cartesia Sonic 3.52026.06 | 2.37 | 2.71 | 75.4 | |
| Inworld TTS 22026.06 | 2.67 | 2.83 | 56.2 | |
| Fish S2 Pro2026.06 | 2.97 | 2.69 | 68.2 | |
| VoxCPM 22026.06 | 3.19 | 2.39 | 70.4 | |
| ZONOS2 8BQuality Mode=false2026.06 | 5.38 | 2.42 | 69.4 | |
| Ground truth2026.06 | 7.34 | 2.22 | 78 |