Zero-Shot Text-to-Speech on YoBind
4.31MOS (Naturalness)Ground Truth (Recordings)
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| Ground Truth (Recordings)Training Data=-, DC=-, OS=-2026.03 | 4.31 | — | |
| Fish-SpeechSource ZS-TTS Model=FS, Training Data=Real 10% + Synth 90%, DC=✓, OS=✓2026.03 | 3.58 | 66.7 | |
| Real-OnlyTraining Data=Real 100%, DC=-, OS=-2026.03 | 3.5 | — | |
| Fish-SpeechSource ZS-TTS Model=FS, Training Data=Real 10% + Synth 90%, DC=-, OS=-2026.03 | 3.5 | — | |
| Real-OnlyTraining Data=Real 10%, DC=-, OS=-2026.03 | 2.85 | — | |
| CosyVoice 2Source ZS-TTS Model=CV2, Training Data=Real 10% + Synth 90%, DC=✓, OS=✓2026.03 | — | 60.2 |