Speech Resynthesis on LibriSpeech clean (test)
4.32UTMOSStable Codec
Evaluation Results
| Method | Links | ||||||
|---|---|---|---|---|---|---|---|
| Stable CodecBitrate (kbps)=0.702026.05 | 4.32 | 4.97 | 94.7 | — | — | — | |
| SylReg-Distill2026.07 | 4.31 | 2.53 | — | 76.8 | — | — | |
| SylReg-DistillChunk size (C)=12026.07 | 4.31 | 2.91 | — | 82.3 | — | — | |
| SylReg-DistillModel=SylReg-Distill, variant=ours2026.07 | 4.31 | — | — | 76.8 | 2.53 | 5.46 | |
| SylReg-DistillModel=SylReg-Distill, Chunk size (C)=12026.07 | 4.31 | — | — | 82.3 | 2.91 | 6.06 | |
| LATTE LARGEBitrate (kbps)=0.652026.05 | 4.23 | 5.82 | 97.4 | — | — | — | |
| LATTE BASEBitrate (kbps)=0.652026.05 | 4.2 | 7.07 | 96.9 | — | — | — | |
| Sylber2026.07 | 4.13 | 1.43 | — | — | — | — | |
| SylberModel=Sylber2026.07 | 4.13 | — | — | — | 1.43 | 3.5 | |
| BigCodecBitrate (kbps)=1.042026.05 | 4.11 | 2.55 | 98.5 | — | — | — | |
| Ground truth2026.07 | 4.1 | 0.72 | — | — | — | — | |
| Ground truthModel=Ground truth2026.07 | 4.1 | — | — | — | 0.72 | 1.97 | |
| Reference2026.05 | 4.09 | 0 | 100 | — | — | — | |
| FocalCodecBitrate (kbps)=0.652026.05 | 4.05 | 2.18 | 97.4 | — | — | — | |
| SylBoostFrequency=6.25Hz2026.07 | 3.89 | 3.16 | — | 75.2 | — | — | |
| SylBoostModel=SylBoost, Frequency=6.25Hz2026.07 | 3.89 | — | — | 75.2 | 3.16 | 6.44 | |
| TWIST2026.07 | 3.85 | 2.59 | — | 174.8 | — | — | |
| TWISTModel=TWIST2026.07 | 3.85 | — | — | 174.8 | 2.59 | 5.65 | |
| WavTokenizerBitrate (kbps)=0.482026.05 | 3.78 | 11.55 | 95.4 | — | — | — | |
| WavLM6-KMBitrate (kbps)=0.452026.05 | 3.75 | 6.2 | 90 | — | — | — | |
| MimiBitrate (kbps)=0.692026.05 | 3.29 | 5.73 | 96 | — | — | — | |
| SemantiCodecBitrate (kbps)=0.652026.05 | 2.91 | 8.97 | 96 | — | — | — | |
| SpeechTokenizerBitrate (kbps)=1.002026.05 | 2.28 | 5.14 | 91.6 | — | — | — | |
| EnCodecBitrate (kbps)=1.502026.05 | 1.58 | 8.08 | 93.8 | — | — | — | |
| DACBitrate (kbps)=1.002026.05 | 1.29 | 20.04 | 89.2 | — | — | — |