Has anyone noticed this? Deepgram nova-2 was great at auto detecting multiple languages, but sometime last week it stopped recognizing Chinese. Anyone knows what is happening?

stt=inference.STT(
    model="deepgram/nova-2",
    language="multi",
    
),

or

stt=inference.STT(
    model="deepgram/nova-3",
    language="multi",
    
),

Both worked before but do not work anymore, and no error messages either.

Wonder if this fix got lost: https://github.com/orgs/deepgram/discussions/1137 Check with Deepgram as they are most likely to know

Deepgram’s “multi” language code code not support Chinese / zh. Nova-2's multi code support English and Spanish. Nova-3 supports 10 langs.