Audio Tokenization and Compression Evaluation on Codec-SUPERB
7.5SDREnCodec
Evaluation results
| Method | Links | |||||||||
|---|---|---|---|---|---|---|---|---|---|---|
| EnCodecFrame rate=75Hz, #Params=15M2025.06 | 7.5 | 1.03 | 2.48 | 70.28 | 2.44 | 3.6 | 7.12 | 87.75 | — | |
| LM-SPTFrame rate=12.5Hz, #Params=74M2025.06 | 5.54 | 0.75 | 2.21 | 73.54 | 2.67 | 3.65 | 5.47 | 56.55 | 79.25 | |
| DualCodecFrame rate=12.5Hz, #Params=665M2025.06 | 4.48 | 0.78 | 2.37 | 74.65 | 2.81 | 3.71 | 5.44 | 55.4 | 77.57 | |
| DACFrame rate=75Hz, #Params=75M2025.06 | 3.62 | 0.68 | 2.98 | 71.88 | 2.33 | 3.42 | 6.63 | 84.45 | — | |
| MimiFrame rate=12.5Hz, #Params=79M2025.06 | 2.88 | 1.17 | 1.87 | 69.24 | 4.82 | 6.24 | 10.14 | 64.45 | 72.7 | |
| SpeechTokenizerFrame rate=50Hz, #Params=103M2025.06 | 1.52 | 0.84 | 1.98 | 74.17 | 3.19 | 4.42 | 6.26 | 50.9 | 74.03 | |
| WavTokenizerFrame rate=40Hz, #Params=81M2025.06 | -9.96 | 1.11 | 1.39 | 60.07 | 17.99 | 20.35 | 24.73 | 23.3 | — | |
| CosyVoice2Frame rate=25Hz, #Params=257M2025.06 | -15.22 | 1.87 | 1.17 | 58.61 | 21.04 | 7.44 | 10.2 | 31.75 | 75.07 |