List your company

Speech Reconstruction on LibriSpeech clean (test)

4.31UTMOS ScoreStableCodec
2.99963.33983.684.0202Sep 26, 2025Nov 10, 2025Dec 26, 2025Feb 10, 2026Mar 28, 2026May 13, 2026Jun 28, 2026

Evaluation results

MethodLinks
StableCodecToken Rate=25, Layers=Single-layer2026.014.315.72.679.30.930.91———————
DTM-Codec@50HzParams=127M, Frame Rate=50, Content=700, Pos=100, Total bps=8002026.064.222.91————782.66—93——3.39
FlexiCodec (τ = 1.0)Params=450M, Frame Rate=12.46, Content=1234, Pos=37, Total bps=12712026.064.22.25————852.82—94——3.2
DTM-Codec@80HzParams=127M, Frame Rate=80, Content=1120, Pos=160, Total bps=12802026.064.22.98————872.95—95——3.42
FlexiCodec (τ = 0.91)Params=450M, Frame Rate=8.26, Content=818, Pos=25, Total bps=8432026.064.192.35————782.46—92——3.17
DTM-Codec@40HzParams=127M, Frame Rate=40, Content=560, Pos=80, Total bps=6402026.064.193.27————742.49—92——3.43
PASTToken Rate=400, Layers=Multi-layer2026.014.182.10.782.40.990.92———————
BiCodecToken Rate=50, Layers=Single-layer2026.014.182.50.9750.980.91———————
Kanade 12.5HzToken Rate=12.5, Layers=Single-layer2026.014.173.31.374.60.970.85———————
Kanade 25HzToken Rate=25, Layers=Single-layer2026.014.162.40.8750.970.88———————
FlexiCodec (τ = 0.867)Params=450M, Frame Rate=6.23, Content=617, Pos=19, Total bps=6362026.064.142.8————712.19—90——3.14
X-Codec 2Token Rate=50, Layers=Single-layer2026.014.132.50.9770.980.9———————
VAETraining Data (hrs)=60002026.054.134.07——0.97————98———
Semantic-VAETraining Data (hrs)=60002026.054.134.07——0.97————98———
X-Codec 2.0Params=210M, Frame Rate=50, Content=800, Pos=0, Total bps=8002026.064.132.57————822.44—92——3.23
Semantic-VAEFrame/s=40, Dim=642026.064.129——————3.968—98.1———
DualCodecToken Rate=100, Layers=Multi-layer2026.014.122.10.775.60.980.95———————
FACodecToken Rate=480, Layers=Multi-layer2026.014.112.10.781.40.980.94———————
BigCodecFrame rate=80Hz, Codebook size(s)=81922026.064.113.6——0.84——2.68—93.5—2.93—
BigCodecParams=159M, Frame Rate=80, Content=1040, Pos=0, Total bps=10402026.064.112.87————842.68—94——3.36
DTM-Codec@25HzParams=127M, Frame Rate=25, Content=350, Pos=50, Total bps=4002026.064.114.73————582.07—90——3.37
SARAFrame/s=50, Dim=642026.064.1——————4.389—99.3———
XCodec2+SGFrame rate=50Hz, Codebook size(s)=655362026.064.13.2——0.8——2.39—91.5—3.41—
Vanilla VAEFrame/s=50, Dim=642026.064.095——————4.076—98.3———
StableToken#C=1, Frame Rate=25Hz, BPS=3252025.094.093.84———————————
Ground Truth2026.054.093.64——1————100———
XCodec2Frame rate=50Hz, Codebook size(s)=81922026.064.094.1——0.72——2.03—89.2—3.84—
Ground Truth2026.064.092.08————1004.64—100——3.23
GT2026.064.086————————————
XCodec2+SGFrame rate=50Hz, Codebook size(s)=81922026.064.083.8——0.73——2.13—89.8—3.6—
Ground Truth2026.014.071.90.678—————————
KM 25HzToken Rate=252026.014.072.7172.40.960.67———————
GLM-4-Voice-Token.#C=1, Frame Rate=12.5Hz, BPS=1752025.094.074.04———————————
XCodec2Frame rate=50Hz, Codebook size(s)=655362026.064.063.2——0.79——2.28—91—3.57—
BigCodecFrame rate=40Hz, Codebook size(s)=81922026.064.056.7——0.66——2.11—89.4—3.72—
KM 12.5HzToken Rate=12.52026.014.0431.172.10.960.66———————
WavCubeTraining Data (hrs)=9602026.054.044.2——0.94————97———
DAC (16k)Params=76M, Frame Rate=50, Content=8000, Pos=0, Total bps=80002026.064.022.14————953.97—97——3.09
WavCube-ProTraining Data (hrs)=60002026.0544.12——0.95————97———
WavTokenizer (70Hz)Params=80.9M, Frame Rate=75, Content=900, Pos=0, Total bps=9002026.0644.32————682.38—91——2.76
TAAE (700 bps)Params=950M, Frame Rate=25, Content=700, Pos=0, Total bps=7002026.063.926.21————572.16—91——3.1
Cont. 50HzToken Rate=50Hz2026.013.920.672.10.990.94———————
STToken Rate=400, Layers=Multi-layer2026.013.92.10.7760.980.92———————
WavTokenizerFrame rate=75Hz, Codebook size(s)=40962026.063.896.8——0.59——2.05—88.6—4—
TS3CodecFrame rate=50Hz, Codebook size(s)=655362026.063.853.6——0.68——2.22—90.9—3.52—
TS3CodecFrame rate=50Hz, Codebook size(s)=1310722026.063.843.6——0.68——2.23—91—3.5—
TAAE (400 bps)Params=950M, Frame Rate=25, Content=400, Pos=0, Total bps=4002026.063.819.39————532—89——3.03
WavTokenizer (40Hz)Params=80.9M, Frame Rate=40, Content=480, Pos=0, Total bps=4802026.063.788.16————571.88—87——3.11
WavTokenizerFrame rate=40Hz, Codebook size(s)=40962026.063.778——0.57——1.88—86.8—4.32—
VARSTok (τ = 0.8)Params=80.9M, Frame Rate=34.52, Content=414, Pos=69, Total bps=4832026.063.7410.51————501.69—86——3.08
TS3CodecFrame rate=40Hz, Codebook size(s)=1310722026.063.734.5——0.63——2.06—89.7—3.75—
TS3CodecFrame rate=40Hz, Codebook size(s)=655362026.063.694.9——0.61——2.01—89.3—3.81—
VARSTok (τ = 0.7)Params=80.9M, Frame Rate=29.03, Content=348, Pos=58, Total bps=4062026.063.6415.39————431.54—84——3.03
Mel-spectrogramTraining Data (hrs)=5852026.053.633.86——0.93————98———
VocosFrame/s=93.75, Dim=1002026.063.625——————3.605—97.7———
VARSTok (τ = 0.6)Params=80.9M, Frame Rate=25.02, Content=300, Pos=50, Total bps=3502026.063.5820.69————371.46—82——3.01
WavTokenizerToken Rate=40, Layers=Single-layer2026.013.579.44.772.10.920.91———————
S3 Tokenizer#C=1, Frame Rate=25Hz, BPS=3002025.093.45.78———————————
CosyVoice2#C=1, Frame Rate=25Hz, BPS=3252025.093.364.25———————————
SNAC (12+23+47 Hz)Params=19.8M, Frame Rate=12+23+47, Content=980, Pos=0, Total bps=9802026.063.054.39————581.91—88——2.84
Baseline (Mel/Fbank)Adaptive Weighting=false2026.04———————3.6—97.884.9——
DACFrame rate=50Hz, Codebook size(s)=1024 × 82026.06————0.87——2.72—94———
DACFrame rate=50Hz, Codebook size(s)=1024 × 22026.06————0.32——1.13—73———
DAS-VAEAdaptive Weighting=false2026.04———————3.98—97.988.8——
DAS-VAEAdaptive Weighting=true2026.04———————2.73—9474.3——
EnCodecAdaptive Weighting=false2026.04———————2.77—93.874.6——
JMAS-VAEAdaptive Weighting=false2026.04———————3.97—97.888.6——
JMAS-VAEAdaptive Weighting=true2026.04———————3.84—97.387.1——
Semantic-VAEAdaptive Weighting=false2026.04———————3.97—98.188.8——
TAS-VAEAdaptive Weighting=false2026.04———————3.97—97.788.6——
TAS-VAEAdaptive Weighting=true2026.04———————2.92—94.776.6——
Vanilla VAEAdaptive Weighting=false2026.04———————4.12—98.590.5——
VevoReference=from [11]2025.11—15————621.150.7421———