The same audio. Why are the results on Hugging Face different from the results of the model on GitHub?

#31
by deepsadness - opened

audio :

result from huggingface :

{'text': '我多謝你封大利是呀真你知收我大利是嗎?你知收咗我錢架嘛?咁你唔好成場鬼晒咁嘈啦我老頭又好大聲嚟講架啦唔係大聲就好戲架楊師傅你尤其是最後嗰句好好味呀嗰句你戒到幾硬幾假盲嘅都知道你做緊戲啦你又有冇職業道德架呢?當時唔係要你整個水煎膠加雙氧水?咁梗係

result from github:

Detected language: zh
新野

I've encountered the same issue. Regarding GitHub, are you utilizing models from
https://github.com/openai/whisper
?

Sign up or log in to comment