Speech Reconstruction on LibriSpeech short splits (4s-10s) (test-clean)
5.12WERAugCodec-3
Evaluation Results
| Method | Links | ||||
|---|---|---|---|---|---|
| AugCodec-3Frame rate=12.5, global, 6.25, Bit-rate (bps)=400.00, Codebook size=16384, 4096, 40962026.06 | 5.12 | 1.99 | 90 | 3.04 | |
| AugCodec-2Frame rate=12.5, global, 6.25, Bit-rate (bps)=387.50, Codebook size=8192, 4096, 40962026.06 | 5.66 | 1.94 | 90 | 2.93 | |
| AugCodec-1Frame rate=12.5, global, 6.25, Bit-rate (bps)=362.50, Codebook size=2048, 4096, 40962026.06 | 5.71 | 1.92 | 90 | 3.11 | |
| AugCodec-2 w/o LaugFrame rate=12.5, global, 6.25, Bit-rate (bps)=387.50, Codebook size=8192, 4096, 40962026.06 | 6.29 | 1.89 | 90 | 2.92 | |
| MimiFrame rate=12.5, Bit-rate (bps)=412.50, Codebook size=2048 × 32026.06 | 7.19 | 1.81 | 92 | 2.25 | |
| Qwen-TTS-Tokenizer-12HzFrame rate=12.5, Bit-rate (bps)=412.50, Codebook size=2048 × 32026.06 | 13.09 | 1.36 | 78 | 1.17 | |
| AugCodec-4Frame rate=6.25, global, 6.25, Bit-rate (bps)=231.25, Codebook size=8192, 4096, 40962026.06 | 17.11 | 1.67 | 88 | 2.78 | |
| BiCodecFrame rate=12.5, global, Bit-rate (bps)=312.50, Codebook size=8192, 40962026.06 | 60.15 | 1.4 | 88 | 2.66 |