Text-to-Speech on Difficult-Word Benchmark English 500 words
0.29D-PERGRAFT
Evaluation results
| Method | Links | ||||||
|---|---|---|---|---|---|---|---|
| GRAFTInput Modality=Audio Hint2026.07 | 0.29 | 0.81 | 92 | 0.09 | 4.47 | 76 | |
| Qwen3-TTS-0.6BInput Modality=Text2026.07 | 0.37 | 1.16 | 91 | 0.02 | 4.46 | 73 | |
| F5-TTSInput Modality=Text2026.07 | 0.38 | 1.1 | 89 | 0.02 | 4.35 | 79 | |
| CosyVoice2Input Modality=Text2026.07 | 0.39 | 1.16 | 90 | 0.03 | 4.46 | 71 | |
| XTTS-v2Input Modality=Text2026.07 | 0.41 | 1.2 | 90 | 0.02 | 4.19 | 67 | |
| Matcha (ph.)Input Modality=Phonemes, Speaker Setting=Single-Speaker2026.07 | 0.43 | 1.11 | 89 | 0.06 | 4.34 | 15 | |
| MaskGCT (ph.)Input Modality=Phonemes2026.07 | 0.46 | 1.15 | 88 | 0.17 | 4.22 | 80 | |
| Piper (ph.)Input Modality=Phonemes, Speaker Setting=Single-Speaker2026.07 | 0.48 | 1.12 | 86 | 0.07 | 4.34 | 12 | |
| StyleTTS2 (ph.)Input Modality=Phonemes2026.07 | 0.65 | 1.11 | 83 | 0.17 | 4.35 | 62 |