SOTA discrete acoustic codec models with 40 tokens per second for audio language modeling - View it on GitHub