whisper identified the wrong language

#41
by lypspeech - opened

When I follow the example of long-form transcription for whisper-large with Korean, the result is English. But after finetuning the whisper-large model with some Korean data, the checkpoint can output Korean. I also test other model size, but all the models output English.
I was confused about it. How should I do to output Korean with the original model?

you can try this:
pipe = pipeline(
"automatic-speech-recognition",
model="openai/whisper-large-v2",
generate_kwargs={"language": "br", "task": "transcribe"},
device="cpu",
use_fast=True
)

You should read the doc about how to properly set the task for transcription instead of translation! As mentioned by @atulyaatul

ArthurZ changed discussion status to closed

Sign up or log in to comment