Waveform Reconstruction on LibriSpeech clean (test)
4.09UTMOSReference audio
Evaluation Results
| Method | Links | ||||||
|---|---|---|---|---|---|---|---|
| Reference audio2026.06 | 4.09 | — | — | — | — | — | |
| OLIVE-JVocoder=HiFi-GAN V2, Integration Type=frozen feat-cond.2026.06 | 3.83 | 0.942 | 3.06 | 4.34 | 2.94 | 4.69 | |
| OLIVE-JVocoder=HiFi-GAN V2, Integration Type=integrated vocoder2026.06 | 3.78 | 0.921 | 2.88 | 4.21 | 2.5 | 4.25 | |
| OLIVE-AVocoder=HiFi-GAN V2, Integration Type=Mix+Gain2026.06 | 3.7 | 0.929 | 2.62 | 4.18 | -3.36 | 1.06 | |
| WavLM BaseVocoder=HiFi-GAN V2, Size=Base2026.06 | 3.68 | 0.917 | 2.21 | 4.05 | -10.93 | -1.31 | |
| HuBERT BaseVocoder=HiFi-GAN V2, Size=Base2026.06 | 3.63 | 0.914 | 2.11 | 4.05 | -11.78 | -1.57 | |
| data2vec 2.0 BaseVocoder=HiFi-GAN V2, Size=Base2026.06 | 3.63 | 0.917 | 2.21 | 4.1 | -9.67 | -1.11 | |
| wav2vec 2.0 BaseVocoder=HiFi-GAN V2, Size=Base2026.06 | 3.58 | 0.912 | 2.19 | 4.04 | -7.53 | -0.42 | |
| data2vec BaseVocoder=HiFi-GAN V2, Size=Base2026.06 | 3.5 | 0.902 | 2.1 | 4.06 | -10.64 | -1.15 | |
| Mel SpectrogramVocoder=HiFi-GAN V22026.06 | 2.98 | 0.833 | 1.42 | 3.59 | -18.02 | -2.6 |