Speech Reconstruction on LibriSpeech (test-clean)
4.23UT MOSStableCodec
Evaluation results
| Method | Links | |||||||||
|---|---|---|---|---|---|---|---|---|---|---|
| StableCodecToken Rate=50, Codebook Size=15625, Disentanglement Ability=false2025.12 | 4.23 | 5.12 | 0.91 | 2.24 | 2.91 | 0.62 | — | — | — | |
| StableCodecNq=2, token/s=502025.12 | 4.23 | — | 0.91 | — | — | — | 2.91 | — | — | |
| X-codecNq=2, token/s=1002025.12 | 4.21 | — | 0.86 | — | — | — | 2.88 | — | — | |
| X-CodecFS=16k, Unified=false, FPS=50, Nq=4, BPS=2000, Paras=178M2025.12 | 4.21 | 3.13 | 0.87 | — | — | — | 2.77 | 0.72 | — | |
| UniAudio-TokenFrame Rate=25Hz, BPS=3252026.05 | 4.19 | 3.47 | — | — | — | — | — | — | — | |
| BiCodecToken Rate=50, Codebook Size=8192, Disentanglement Ability=false2025.12 | 4.18 | — | 0.92 | 2.51 | 3.13 | 0.8 | — | — | — | |
| BiCodecFS=16k, Unified=false, FPS=50, Nq=1, BPS=650, Paras=156M2025.12 | 4.18 | 3.23 | 0.92 | — | — | — | 2.51 | 0.8 | — | |
| MSR-Codec-424Token Rate=62.5, Codebook Size=500/32/64, Disentanglement Ability=true2025.12 | 4.15 | — | 0.84 | 2.37 | 1.82 | 0.8 | — | — | — | |
| X-codec2Token Rate=50, Codebook Size=65536, Disentanglement Ability=false2025.12 | 4.13 | 2.47 | 0.92 | 2.43 | 3.04 | 0.82 | — | — | — | |
| X-codec2Nq=1, token/s=502025.12 | 4.13 | — | 0.92 | — | — | — | 3.04 | — | — | |
| X-Codec2FS=16k, Unified=false, FPS=50, Nq=1, BPS=800, Paras=242M2025.12 | 4.13 | 3.53 | 0.92 | — | — | — | 2.43 | 0.82 | — | |
| FlexiCodecFS=16k, Unified=false, FPS=8.15, Nq=8, BPS=831, Paras=216M2025.12 | 4.12 | 2.79 | 0.92 | — | — | — | 2.46 | 0.81 | — | |
| BigCodecToken Rate=80, Codebook Size=8192, Disentanglement Ability=false2025.12 | 4.11 | 2.76 | 0.93 | 2.68 | 3.27 | 0.84 | — | — | — | |
| BigCodecNq=1, token/s=802025.12 | 4.11 | — | 0.93 | — | — | — | 3.27 | — | — | |
| DisCodecToken Rate=50, Codebook Size=65536, Disentanglement Ability=true2025.12 | 4.1 | 2.92 | 0.86 | 1.98 | 2.33 | 0.81 | — | — | — | |
| Ground TruthDisentanglement Ability=false2025.12 | 4.09 | 1.96 | 1 | 4.64 | 4.55 | 1 | — | — | — | |
| StableTokenFrame Rate=25Hz, BPS=3252026.05 | 4.09 | 3.84 | — | — | — | — | — | — | — | |
| H-Codec-1.0FS=16k, Unified=true, FPS=25, Nq=4+4, BPS=2000, Paras=146M2025.12 | 4.08 | 2.85 | 0.94 | — | — | — | 3.05 | 0.89 | — | |
| Q2D2Nq=1, token/s=3332025.12 | 4.07 | — | 0.96 | — | — | — | 3.79 | — | — | |
| Q2D2Nq=1, token/s=1662025.12 | 4.07 | — | 0.95 | — | — | — | 3.36 | — | — | |
| GLM-4-Voice-TokenizerFrame Rate=12.5Hz, BPS=1752026.05 | 4.07 | 4.04 | — | — | — | — | — | — | — | |
| X-codecToken Rate=50, Codebook Size=1024, Disentanglement Ability=false2025.12 | 4.05 | 3.42 | 0.83 | 1.84 | 2.38 | 0.52 | — | — | — | |
| X-codecNq=1, token/s=502025.12 | 4.05 | — | 0.83 | — | — | — | 2.38 | — | — | |
| Q2D2Nq=1, token/s=662025.12 | 4.04 | — | 0.92 | — | — | — | 2.5 | — | — | |
| DACNq=12, token/s=6002025.12 | 4 | — | 0.95 | — | — | — | 4.15 | — | — | |
| WavTokenizerFS=24k, Unified=false, FPS=75, Nq=1, BPS=900, Paras=77M2025.12 | 4 | 5.43 | 0.91 | — | — | — | 2.39 | 0.68 | — | |
| UniCodecFS=24k, Unified=true, FPS=75, Nq=1, BPS=900, Paras=274M2025.12 | 4 | 4.23 | 0.92 | — | — | — | 2.56 | 0.76 | — | |
| H-Codec-1.5FS=16k, Unified=false, FPS=19.54, Nq=4+4, BPS=1622, Paras=678M2025.12 | 3.99 | 3.02 | 0.92 | — | — | — | 2.6 | 0.83 | — | |
| H-Codec-2.0 (Large)FS=48k, Unified=true, FPS=6.25, Nq=16+16, BPS=2000, Paras=1.2B2025.12 | 3.89 | 2.7 | 0.94 | — | — | — | 3.1 | 0.92 | — | |
| X-CodecFS=16k, Unified=true, FPS=50, Nq=4, BPS=2000, Paras=178M2025.12 | 3.88 | 3.33 | 0.92 | — | — | — | 2.64 | 0.77 | — | |
| WavTokenizerToken Rate=75, Codebook Size=4096, Disentanglement Ability=false2025.12 | 3.79 | 3.98 | 0.9 | 2.13 | 2.63 | 0.65 | — | — | — | |
| WavTokenizerNq=1, token/s=752025.12 | 3.79 | — | 0.9 | — | — | — | 2.63 | — | — | |
| WavTokenizerFS=24k, Unified=true, FPS=40, Nq=1, BPS=480, Paras=77M2025.12 | 3.78 | 10.03 | 0.87 | — | — | — | 1.88 | 0.57 | — | |
| H-Codec-2.0 (Small)FS=48k, Unified=true, FPS=6.25, Nq=16+16, BPS=2000, Paras=236M2025.12 | 3.75 | 3.05 | 0.93 | — | — | — | 2.67 | 0.89 | — | |
| WavTokenizerNq=1, token/s=402025.12 | 3.57 | — | 0.85 | — | — | — | 2.06 | — | — | |
| MimiNq=8, token/s=1002025.12 | 3.56 | — | 0.91 | — | — | — | 2.8 | — | — | |
| MagiCodecStreaming/Non-streaming mode=Streaming2026.03 | 3.4901 | 4.35 | — | — | — | 0.9715 | — | — | 1.85 | |
| MimiNq=6, token/s=752025.12 | 3.38 | — | 0.89 | — | — | — | 2.51 | — | — | |
| WavTokenizerFrame Rate=75Hz, BPS=9002026.05 | 3.37 | 5.07 | — | — | — | — | — | — | — | |
| BigCodecStreaming/Non-streaming mode=Non-Streaming2026.03 | 3.3694 | 3.67 | — | — | — | 0.9799 | — | — | 1.5 | |
| CosyVoice2Frame Rate=25Hz, BPS=3252026.05 | 3.36 | 4.25 | — | — | — | — | — | — | — | |
| TAAEStreaming/Non-streaming mode=Non-Streaming2026.03 | 3.3495 | 8.78 | — | — | — | 0.9371 | — | — | 4.38 | |
| JHCodec-M-8Streaming/Non-streaming mode=Streaming2026.03 | 3.3229 | 3.19 | — | — | — | 0.9826 | — | — | 1.25 | |
| Ground Truth (GT)Mode=Ground Truth2026.03 | 3.2311 | 2.99 | — | — | — | 1 | — | — | 1.13 | |
| NanoCodecStreaming/Non-streaming mode=Non-Streaming2026.03 | 3.163 | 3.16 | — | — | — | 0.9886 | — | — | 1.24 | |
| EncodecNq=8, token/s=6002025.12 | 3.09 | — | 0.94 | — | — | — | 3.18 | — | — | |
| EncodecFS=24k, Unified=false, FPS=75, Nq=8, BPS=6000, Paras=15M2025.12 | 3.09 | 2.64 | 0.94 | — | — | — | 2.77 | 0.89 | — | |
| MimiToken Rate=50, Codebook Size=2048, Disentanglement Ability=false2025.12 | 3.03 | 4.89 | 0.85 | 1.64 | 2.09 | 0.5 | — | — | — | |
| MimiNq=4, token/s=502025.12 | 3.03 | — | 0.85 | — | — | — | 2.09 | — | — | |
| FocalCodec-StreamStreaming/Non-streaming mode=Streaming2026.03 | 2.9772 | 4.05 | — | — | — | 0.9606 | — | — | 1.66 | |
| Mimi-32Streaming/Non-streaming mode=Streaming2026.03 | 2.9685 | 3.26 | — | — | — | 0.9898 | — | — | 1.29 | |
| Mimi-8Streaming/Non-streaming mode=Streaming2026.03 | 2.7884 | 4.07 | — | — | — | 0.9673 | — | — | 1.78 | |
| SemantiCodecNq=2, token/s=502025.12 | 2.71 | — | 0.84 | — | — | — | 2.18 | — | — | |
| DAC-8Streaming/Non-streaming mode=Non-Streaming2026.03 | 2.5845 | 3.33 | — | — | — | 0.9832 | — | — | 1.29 | |
| SpeechTokenizerNq=2, token/s=1002025.12 | 2.28 | — | 0.77 | — | — | — | 1.59 | — | — | |
| DACFS=16k, Unified=true, FPS=50, Nq=4, BPS=2000, Paras=76M2025.12 | 1.83 | 4.32 | 0.84 | — | — | — | 1.42 | 0.6 | — | |
| EncodecNq=2, token/s=1502025.12 | 1.58 | — | 0.85 | — | — | — | 1.94 | — | — | |
| DACNq=2, token/s=1002025.12 | 1.29 | — | 0.73 | — | — | — | 1.4 | — | — | |
| SpeechTokenizerToken Rate=50, Codebook Size=1024, Disentanglement Ability=false2025.12 | 1.27 | 5.01 | 0.64 | 1.14 | 1.3 | 0.17 | — | — | — | |
| SpeechTokenizerNq=1, token/s=502025.12 | 1.27 | — | 0.64 | — | — | — | 1.3 | — | — | |
| EncodecToken Rate=75, Codebook Size=1024, Disentanglement Ability=false2025.12 | 1.25 | 28.92 | 0.77 | 1.23 | 1.48 | 0.25 | — | — | — | |
| DACToken Rate=50, Codebook Size=1024, Disentanglement Ability=false2025.12 | 1.25 | 74.55 | 0.62 | 1.06 | 1.2 | 0.08 | — | — | — | |
| EncodecNq=1, token/s=752025.12 | 1.25 | — | 0.77 | — | — | — | 1.48 | — | — | |
| DACNq=1, token/s=502025.12 | 1.25 | — | 0.62 | — | — | — | 1.2 | — | — |