...
Build, evaluate, or fine-tune Cantonese (粤语 / Yue) speech systems using the WenetSpeech-Yue ecosystem from ASLP@NPU: a 21,800-hour annotated Cantonese corpus, the WenetSpeech-Pipe preprocessing pipeline, the WSYue-eval ASR/TTS benchmarks, and the released Conformer-Yue / Whisper-m-Yue / SenseVoice-s-Yue ASR models plus CosyVoice2-Yue and Llasa-1B-Yue-Updated zero-shot TTS models. Use this skill to choose the right sub-component for a Cantonese speech task (data prep, ASR inference/finetune, TTS synthesis, benchmark evaluation) and to follow the official repo's published recipes end-to-end. Triggers: cantonese speech, 粤语语音, yue ASR, cantonese TTS, cantonese corpus, WenetSpeech-Yue, WSYue-eval, Conformer-Yue, CosyVoice2-Yue, Llasa-Yue.
Created by: songlin she · GitHub