Spaces:

openai

/

whisper

like 1.56k

Running on T4

There's a bug/misbehavior to report of the way it processes Cantonese language. Whisper seems to work for Cantonese, but it actually generates Cantonese language and "translates" it into Mandarin texts, which is improper handling of Cantonese speech.

Screenshot below shows the work it done:

It's actually an approach to translating Cantonese language into another, rather than doing what it is supposed to do - dictating - to convert speech into text ONLY.

Please correct the mistake and let it listen to spoken Cantonese and output written Cantonese.

Thank you!