Audio tokenizer and detokenizer designed for speech LLMs. Underpins the LongCat-Flash-Omni model's audio capabilities.

Paper

audio

Related