Ming-3.0-asr-tiny

About

Ming-3.0-asr-tiny is a speech recognition model that pairs the Ling-tiny-v3 language backbone with an audio encoder. It transcribes single-turn Chinese and English speech, was evaluated on 15 regional dialect test sets, and can take hotwords or preceding dialogue in the prompt to better recognize domain-specific terms. The repository includes weights, tokenizer, a custom Transformers implementation and a runnable example.