Subjective Speech Synthesis Evaluation on Seed-TTS and CosyVoice en 3 (test)
0CMOSMI-DFM (GibbsTTS)
Evaluation results
| Method | Links | ||
|---|---|---|---|
| MI-DFM (GibbsTTS)Scheduler=Numerical KO2026.05 | 0 | 4.18 | |
| Masked DFMScheduler=Closed-form KO2026.05 | -0.229 | 3.94 | |
| MI-DFMScheduler=Grid-searched2026.05 | -0.257 | 4.05 | |
| Masked DFMScheduler=DiFlow-TTS2026.05 | -0.286 | 3.91 | |
| MI-DFM w/o correctorScheduler=Numerical KO2026.05 | -0.362 | 3.86 | |
| Masked DDScheduler=MaskGCT2026.05 | -0.743 | 3.56 | |
| MaskGCT (original)Scheduler=MaskGCT2026.05 | -0.762 | 4.06 |