Asking for recommendations on recognizing Thai and other languages with FunASR MLT Nano 2512

[edward15241]

Thank you very much for releasing the model and making it publicly available.
I really appreciate your work.

I am currently working on a multilingual ASR task and encountered a small issue when using the MLT model.

I noticed that the MLT model supports 31 languages. However, when I set the language option to auto and run recognition on the GigaSpeech Thai test set, the output contains a large amount of Chinese text. The behavior seems somewhat unstable.

I also tried explicitly changing the language prompt in model.py, but the recognition results are quite similar.

Are there any recommendations or best practices for using the model in multilingual scenarios?

[LauraGPT]

Ok, we would write the detail usage sooner.