Audio codec for Speech LLMs: tokens and streaming
Published on October 17, 2025, the briefing reports the open-source release of an audio tokenizer and detokenizer optimized for Speech LLMs, with parallel tokens and a low-latency streaming decoder.
Source: LongCat-Audio-Codec para Speech LLMs (github.com). Text prepared with AI from this source.
What happened and what to do
On October 17, 2025, a briefing about LongCat-Audio-Codec reports that Meituan released an open-source audio tokenizer and detokenizer for Speech LLMs. It describes parallel semantic and acoustic tokens at 16.7 Hz, low-bitrate encoding, and a low-latency streaming decoder. To consult and verify the details, look up the briefing under the title “LongCat-Audio-Codec para Speech LLMs” and check the original Meituan post it cites.
A company evaluating voice applications can test this approach with representative samples, measure quality, bitrate, and latency, and monitor results through an evaluation pipeline. A deployment decision should also account for integration with existing systems and the product’s operational requirements.
How the consultancy can help
Wendelmaques can diagnose the requirements and risks of a voice application, define an evaluation scope, and implement the integration, testing, and monitoring needed on the company’s infrastructure.
Next step
Send a brief description of the voice application and the systems involved to receive a proposal scoped for diagnosis and implementation.
Consulting for your project
Infrastructure review, deployment and ongoing operations, with scope and pricing defined in the proposal.
Quoted per project
Request a proposal