MOSS-Audio-Tokenizer: Scaling Audio Tokenizers for Future Audio Foundation Models
Open paper
Benchmark
Model
Rank
Results
audio-compression-on-librispeech-test-clean-codec-reconstruction-low-bitrate-0-75-1-5-kbps
MOSS-Audio-Tokenizer (1.0 kbps, 12.5 tokens/s, 8 RVQ)
#1
PESQ-NB: 3.38
SIM: 0.88
STOI: 0.94
PESQ-WB: 2.87
Rank counts only results with a code link.