Text-to-Speech on ZTTS1-Eval In-The-Wild ko
2.27WERGemini 3.1 Flash
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| Gemini 3.1 Flashsupports zero-shot voice cloning=false2026.06 | 2.27 | 3.45 | 29.2 | |
| Qwen 3 TTS 1.7B2026.06 | 2.8 | 2.97 | 73.7 | |
| Cartesia Sonic 3.52026.06 | 3.38 | 2.86 | 77.8 | |
| Fish S2 Pro2026.06 | 3.48 | 2.76 | 72.8 | |
| Eleven Labs V3supports zero-shot voice cloning=false2026.06 | 3.93 | 3.4 | 28.7 | |
| Inworld TTS 22026.06 | 4.16 | 2.86 | 62.6 | |
| ZONOS2 8B Quality ModeQuality Mode=true2026.06 | 4.59 | 2.78 | 67.3 | |
| VoxCPM 22026.06 | 4.89 | 2.51 | 74.8 | |
| ZONOS2 8BQuality Mode=false2026.06 | 7.45 | 2.55 | 72.1 | |
| Ground truth2026.06 | 8.86 | 2.48 | 79.7 |