Speech Synthesis on Audio Turing Test (ATT)
51.5Mean ATT ScoreFish Audio S2
Evaluation results
| Method | Links | |||
|---|---|---|---|---|
| Fish Audio S2instruction expansion=true2026.03 | 51.5 | 6.1 | 51 | |
| Fish Audio S2instruction expansion=false2026.03 | 48.3 | 6.8 | 47.7 | |
| Fish Audio S12026.03 | 47.9 | 8.7 | 47.1 | |
| Seed-TTS2026.03 | 41.7 | 1.1 | 39.8 | |
| MiniMax-Speech2026.03 | 38.7 | 1.1 | 36.8 | |
| Step-Audio2026.03 | 28.6 | 1.1 | 26.6 | |
| CosyVoice2026.03 | 23.4 | 1 | 21.4 | |
| GPT-4o2026.03 | 13.8 | 1.1 | 11.8 |