Voice Conversion on VCTK (test)
4.26nMOSGround truth
Evaluation results
| Method | Links | |||||||
|---|---|---|---|---|---|---|---|---|
| Ground truth2026.02 | 4.26 | 3.62 | 4.14 | 3.75 | 4.05 | 10 | 0.94 | |
| MeanVoiceFlowNFE=12026.02 | 3.87 | 2.92 | 3.98 | 3.78 | 4.1 | 120 | 0.886 | |
| FastVoiceGrad+NFE=1, distillation and adversarial training=true, pretrained teacher and feature extractor=true2026.02 | 3.81 | 2.99 | 3.99 | 3.79 | 4.03 | 120 | 0.888 | |
| VoiceGrad-DMNFE=302026.02 | 3.79 | 2.7 | 3.88 | 3.77 | 4.02 | 140 | 0.885 | |
| VoiceGrad-FMNFE=302026.02 | 3.79 | 2.92 | 3.88 | 3.79 | 4.05 | 110 | 0.885 | |
| FastVoiceGradNFE=1, distillation and adversarial training=true, pretrained teacher and feature extractor=true2026.02 | 3.73 | 2.93 | 3.96 | 3.77 | 4.04 | 130 | 0.888 | |
| DiffVCNFE=302026.02 | 3.59 | 2.55 | 3.76 | 3.75 | 4.03 | 540 | 0.88 | |
| VoiceGrad-FMNFE=12026.02 | 3.14 | 2.6 | 3.81 | 3.69 | 4.01 | 110 | 0.885 | |
| VoiceGrad-DMNFE=12026.02 | 2.77 | 2.48 | 3.72 | 3.68 | 3.99 | 140 | 0.884 |