Language code for Taiwanese transcriptions

[meg261995]

Does whisper support Taiwanese transcriptions? If so, Could someone help me out with what language code to use here? Chinese/Mandarin/Cantonese?

[glangford]

See

• Spoken Chinese languages #25

The full set of language codes supported by whisper is here:

whisper/whisper/tokenizer.py

Line 10 in ba3f3cd

LANGUAGES = {

[jake1271]

Chinese is a bit of a mess on Whisper which seems to randomly mix simplified and traditional. You can try adding a prompt to request simplified Mandarin. Hopefully they'll release a new language code for traditional Mandarin in the future.

[laubonghaudoi]

@meg261995 By Taiwanese do you mean Southern Min in Taiwan, code nan-tw? Whisper doesn't support that at the moment.

[kalvinchang]

Empirically, ChatGPT (which uses Whisper) seems to support Taiwanese but translates it to Mandarin(?

I asked in Taiwanese:

lí thiann ū bô guá teh kóng sánn
你聽有無我咧講啥