Speech Synthesis on LibriTTS
4.5PESQGround Truth
Evaluation Results
| Method | Links | ||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|
| Ground Truth2025.12 | 4.5 | 0 | — | — | — | — | — | — | — | — | |
| PeriodWaveNFE=16, #Param. (M)=29.81, #MACs (Giga/5s)=4986.562026.03 | 4.24 | — | — | — | 3.652 | 4.749 | — | — | — | — | |
| RNDVoC-sharedNFE=1, #Param. (M)=3.14, #MACs (Giga/5s)=37.202026.03 | 4.226 | — | — | — | 3.657 | 4.915 | — | — | — | — | |
| RNDVoC-nonsharedNFE=1, #Param. (M)=9.48, #MACs (Giga/5s)=28.292026.03 | 4.085 | — | — | — | 3.607 | 4.873 | — | — | — | — | |
| PriorGradNFE=50, #Param. (M)=2.70, #MACs (Giga/5s)=83642026.03 | 4.017 | — | — | — | 3.329 | 4.737 | — | — | — | — | |
| WaveFMNFE=6, #Param. (M)=19.50, #MACs (Giga/5s)=11882026.03 | 3.954 | — | — | — | 3.191 | 4.943 | — | — | — | — | |
| ComVo2026.03 | 3.8239 | — | — | 4.07 | 3.6901 | — | 0.8439 | 0.0903 | 96.09 | 0 | |
| FreGradNFE=50, #Param. (M)=1.82, #MACs (Giga/5s)=15892026.03 | 3.793 | — | — | — | 3.217 | 4.699 | — | — | — | — | |
| Vocos2026.03 | 3.6266 | — | — | 4.05 | 3.6025 | — | 0.8856 | 0.1061 | 95.22 | 0.02 | |
| BigVGAN2026.03 | 3.6122 | — | — | 4.05 | 3.5197 | — | 0.8994 | 0.1181 | 94.18 | 0.05 | |
| HiFiGAN + UnivHDVocoder=HiFiGAN, Discriminator=UnivHD (+H)2025.12 | 3.13 | 2.88 | 37.85 | — | — | — | — | — | — | — | |
| iSTFTNet + UnivHDVocoder=iSTFTNet, Discriminator=UnivHD (+H)2025.12 | 3.07 | 2.92 | 40.03 | — | — | — | — | — | — | — | |
| HiFiGAN + MS-SB-CQTVocoder=HiFiGAN, Discriminator=MS-SB-CQT (+C)2025.12 | 3.06 | 2.92 | 39.46 | — | — | — | — | — | — | — | |
| HiFiGAN + MS-STFTVocoder=HiFiGAN, Discriminator=MS-STFT (+S)2025.12 | 3.02 | 2.95 | 41.06 | — | — | — | — | — | — | — | |
| iSTFTNet + MS-SB-CQTVocoder=iSTFTNet, Discriminator=MS-SB-CQT (+C)2025.12 | 3.01 | 2.94 | 41.03 | — | — | — | — | — | — | — | |
| HiFiGANVocoder=HiFiGAN, Discriminator=Original2025.12 | 3 | 2.99 | 46.25 | — | — | — | — | — | — | — | |
| iSTFTNet + MS-STFTVocoder=iSTFTNet, Discriminator=MS-STFT (+S)2025.12 | 2.99 | 2.96 | 40.46 | — | — | — | — | — | — | — | |
| iSTFTNetVocoder=iSTFTNet, Discriminator=Original2025.12 | 2.95 | 2.96 | 43.41 | — | — | — | — | — | — | — | |
| HiFi-GAN2026.03 | 2.936 | — | — | 4 | 3.3453 | — | 1.0455 | 0.1554 | 91.74 | 0.09 | |
| iSTFTNetImplementation=open-source2026.03 | 2.8136 | — | — | 3.98 | 3.3591 | — | 1.1046 | 0.1476 | 92.43 | 0.04 | |
| GT2026.03 | — | — | — | 4.08 | 3.8712 | — | — | — | — | 0.14 |