MMS - is automatic language detection a thing with MMS? The mms_infer.py requires a language input to work

[altryne]

What is your question?

How to auto-detect a language from the mms_infer.py?

Code

Running

python asr/infer/mms_infer.py --model mms1b_l1107.pt --audio output_audio.mp3

shows

the following arguments are required: --lang

[tanmayh-fg]

i have same issue. If not automatic, is there at least a way I can give multiple language inputs, eg eng & hin.

[androstj]

We currently don't support automatic language detection given a speech. But you could use the LID to detect the language and put the predicted language code into inference script (but check if ASR support those languages or not from README table)

[altryne]

We currently don't support automatic language detection given a speech. But you could use the LID to detect the language and put the predicted language code into inference script (but check if ASR support those languages or not from README table)

Thanks for the comment @androstj , could you elaborate on what "use the LID" means?

[androstj]

LID is Language Identification. Given a speech input, LID will predict what language is it. Check our readme section on how to run LID MMS LID