facebook

/

mms-1b-all

like 86

The inference API performs poorly on transcribing because it does not know the language being spoken. If you would allow the target language to be one of the post requests data, I think the model can have better transcription results by effectively using the appropriate language adapter. Please do consider this.