List your company

Speech Reconstruction on LibriTTS clean (test)

4.644PESQGT
1.0375921.9738712.910153.846429Feb 27, 2025May 20, 2025Aug 10, 2025Nov 1, 2025Jan 22, 2026Apr 14, 2026Jul 6, 2026

Evaluation results

MethodLinks
GT2026.034.6441—4.056————————
BigVGANUnified=true, SR=44.1, Token rate category=Vocoder2026.014.1860.993——————————
DACBandwidth=9.0kbps, Nq=9, token/s=9002025.123.90820.9699—3.909797.81———————
DACNq=9, FPS=9002026.033.9080.97—3.91————————
DACUnified=true, TPS=900, kbps=9, Nq=9, SR=44.1, Token rate category=> 1002026.013.90.972——————————
Q2D2Bandwidth=6.9kbps, Nq=1, token/s=3332025.123.70060.9637—4.032197.99———————
DACParams [M]=74.71, Token Rate=75, Codebook Size=1024, Q=8, Bitrate [kbps]=62025.123.52150.95460.9713.6905————————
DACUnified=true, DR=40, TPF=8, TPS=6002025.023.51970.97090.95463.6905————————
Q2D2Bandwidth=3.3kbps, Nq=1, token/s=1662025.123.36350.9557—4.061396.76———————
Ours Wav-VAENq=1, FPS=11.722026.033.2370.967—4.013————————
Ours Wav-VAENq=1, FPS=7.812026.033.0890.963—4.116————————
VibeVoiceNq=1, FPS=7.502026.033.0680.828—4.181————————
UniCodecUnified=true, DR=320, TPF=1, TPS=752025.023.02660.94930.94863.9873————————
HiFi-CodecBandwidth=3.0kbps, Nq=4, token/s=3002025.123.01160.9446—3.903595.76———————
HiFi-CodecBandwidth=4.0kbps, Nq=4, token/s=4002025.122.96110.9405—3.752996.17———————
TIREParams=63.33 M2026.072.9560.944————4.382——0.699—0.728
Dual-TIRE (baseline)Params=63.91 M, segment selection strategy=random (unconstrained)2026.072.9310.946————4.387——0.699—0.727
Dual-TIRE (cross-file)Params=63.91 M, segment selection strategy=cross-file (same speaker and book)2026.072.9230.946————4.41——0.721—0.741
UniCodecUnified=true, TPS=75, kbps=1.3, Nq=1, SR=24, Token rate category=≤ 1002026.012.870.94——————————
EncodecUnified=true, TPS=600, kbps=6, Nq=8, SR=24, Token rate category=> 1002026.012.8190.943——————————
VocosNq=8, FPS=6002026.032.8070.943—3.695————————
VocosBandwidth=6.0kbps, Nq=8, token/s=6002025.122.80690.9426—3.695494.37———————
No-TIRE (baseline)Params=63.12 M2026.072.7970.938————4.362——0.703—0.772
DACBandwidth=4.0kbps, Nq=4, token/s=4002025.122.73780.928—3.432995.72———————
UniSRCodec-LUnified=true, TPS=176, kbps=2.29, Nq=1, SR=44.1, Token rate category=> 1002026.012.7270.941——————————
EncodecUnified=true, DR=40, TPF=8, TPS=6002025.022.72020.93910.95273.0399————————
EncodecParams [M]=14.85, Token Rate=75, Codebook Size=1024, Q=8, Bitrate [kbps]=62025.122.72020.93910.95273.0399————————
EncodecBandwidth=6.0kbps, Nq=8, token/s=6002025.122.72020.9391—3.039995.27———————
EncodecNq=8, FPS=6002026.032.720.939—3.04————————
BigCodecUnified=false, TPS=80, kbps=1.04, Nq=1, SR=16, Token rate category=≤ 1002026.012.70.943——————————
BigCodecNq=1, FPS=802026.032.6970.939—4.097————————
SACodecParams [M]=75.17, Token Rate=75, Codebook Size=1000 / 1024, Q=2, Bitrate [kbps]=1.52025.122.69370.93170.93814.0373————————
BigCodecUnified=false, DR=200, TPF=1, TPS=802025.022.68720.92930.9484.0367————————
SpeechTokenizerUnified=false, DR=40, TPF=8, TPS=6002025.022.61210.91650.94953.8794————————
SpeechTokenizerParams [M]=103.68, Token Rate=50, Codebook Size=1024, Q=8, Bitrate [kbps]=42025.122.61210.91650.94953.8794————————
SpeechTokenizerBandwidth=6.0kbps, Nq=8, token/s=6002025.122.61210.9165—3.879494.95———————
WT-SpeechUnified=false, TPS=75, kbps=0.9, Nq=1, SR=24, Token rate category=≤ 1002026.012.5660.922——————————
SNACUnified=true, TPS=240, kbps=2.88, Nq=4, SR=44.1, Token rate category=> 1002026.012.5610.925——————————
Q2D2Bandwidth=1kbps, Nq=1, token/s=752025.122.50910.9217—4.052694.4———————
WavTokenizerParams [M]=80.55, Token Rate=75, Codebook Size=4096, Q=1, Bitrate [kbps]=0.92025.122.46870.91940.93943.9687————————
VocosBandwidth=3.0kbps, Nq=4, token/s=3002025.122.40260.9231—3.53993.58———————
Wavtokenizer (speech)Unified=false, DR=320, TPF=1, TPS=752025.022.3730.91390.93824.0486————————
WavTokenizerBandwidth=0.9kbps, Nq=1, token/s=752025.122.3730.9139—4.048693.82———————
WavTokenizerNq=1, FPS=752026.032.3730.914—4.049————————
MimiUnified=true, DR=240, TPF=8, TPS=1002025.022.26950.91180.9123.5731————————
FACodecParams [M]=374.49, Token Rate=80, Codebook Size=1024, Q=6, Bitrate [kbps]=4.82025.122.25320.91270.94023.4454————————
EncodecUnified=true, TPS=300, kbps=3, Nq=4, SR=24, Token rate category=> 1002026.012.1160.904——————————
EncodecBandwidth=3.0kbps, Nq=4, token/s=3002025.122.05170.9007—2.30791.98———————
SpeechTokenizerBandwidth=3.0kbps, Nq=4, token/s=3002025.121.93110.8778—3.563292.73———————
WT-UnifiedUnified=true, TPS=40, kbps=0.48, Nq=1, SR=24, Token rate category=≤ 1002026.011.9120.875——————————
TAAEUnified=false, DR=320, TPF=2, TPS=502025.021.89550.88160.9264.1389————————
EG-GRVQVQ scheme (Structure)=EG-GRVQ (2x2), Bitrate=0.6875 kbps2026.031.8810.89———-7.3092.496—————
Mimi (official)VQ scheme (Structure)=RVQ (1x4), Bitrate=0.6875 kbps2026.031.8720.876———3.4512.01—————
Mimi (GRVQ)VQ scheme (Structure)=GRVQ (2x2), Bitrate=0.6875 kbps2026.031.8520.889———-7.2942.464—————
Wavtokenizer (unified)Unified=true, DR=320, TPF=1, TPS=752025.021.83790.87180.91753.6115————————
UniSRCodec-BUnified=true, TPS=40, kbps=0.52, Nq=1, SR=44.1, Token rate category=≤ 1002026.011.8360.875——————————
MelCapUnified=true, TPS=260, kbps=3.4, Nq=1, SR=44.1, Token rate category=> 1002026.011.8020.888——————————
TAAEUnified=false, TPS=50, kbps=0.7, Nq=1, SR=16, Token rate category=≤ 1002026.011.7870.89——————————
Mimi (retrain)VQ scheme (Structure)=RVQ (1x4), Bitrate=0.6875 kbps2026.031.7790.886———-6.9692.546—————
WT-MAUnified=false, TPS=75, kbps=0.9, Nq=1, SR=24, Token rate category=≤ 1002026.011.7470.857——————————
WavTokenizerBandwidth=0.5kbps, Nq=1, token/s=402025.121.70270.8615—3.601691.73———————
EncodecParams [M]=14.85, Token Rate=75, Codebook Size=1024, Q=2, Bitrate [kbps]=1.52025.121.53980.84620.84961.5551————————
DACParams [M]=74.71, Token Rate=75, Codebook Size=1024, Q=2, Bitrate [kbps]=1.52025.121.530.84530.89571.9152————————
DACUnified=true, TPS=100, kbps=1, Nq=1, SR=44.1, Token rate category=≤ 1002026.011.3080.763——————————
SpeechTokenizerParams [M]=103.68, Token Rate=50, Codebook Size=1024, Q=3, Bitrate [kbps]=1.52025.121.26680.78530.83532.4121————————
DACBandwidth=1.0kbps, Nq=1, token/s=1002025.121.24640.7706—1.49479.41———————
DACUnified=false, DR=320, TPF=1, TPS=752025.021.17630.77390.7561.3531————————
BiCodecToken Rate (t/s)=502026.06———4.25—26.49—2.61.30.780.55—
CleanCodec@12.5Token Rate (t/s)=12.52026.06———4.32—21.96—2.71.40.860.44—
CleanCodec@31.25Token Rate (t/s)=31.252026.06———4.27—22.02—1.70.90.880.33—
CleanCodec@62.5Token Rate (t/s)=62.52026.06———4.23—22.94—1.30.60.90.29—
FocalCodecToken Rate (t/s)=12.52026.06———4.21—25.19—8.34.50.450.78—
Ground Truth2025.02———4.0562————————
GT2025.12———4.0562————————
GTToken Rate (t/s)=–2026.06———4.2—26.49—0010—
KanadeToken Rate (t/s)=12.52026.06———4.22—23.6—42.10.650.57—
MimiToken Rate (t/s)=1002026.06———3.86—17.48—3.51.90.720.56—
Qwen3Token Rate (t/s)=2002026.06———4.27—27.99—1.30.70.90.29—
Qwen3Token Rate (t/s)=1002026.06———3.92—19.45—2.61.40.720.38—
Qwen3Token Rate (t/s)=502026.06———2.65—9.01—7.94.50.310.63—
WavTokenizerToken Rate (t/s)=402026.06———3.81—19.98—950.620.46—
XCodec2Token Rate (t/s)=502026.06———4.2—23.51—2.61.50.770.39—