Transcription in different languages for Punjabi audio
Input - Punjabi Audio clip of < 10 seconds
Expected - Transcription in Punjabi (Gurmukhi) alphabet
Observed - Transcription in random languages - English, Urdu, but not Punjabi (Gurmukhi)
Question - Can we fix this on whisper 3, so it shows Punjabi transcription in Gurmukhi alphabets ?
Follow up Question
I fine tuned whisper 2 on my own dataset - input - Gurmukhi sentence, Punjabi audio
I observed similar behavior on fine tuned model based on whisper 2
Do I need to pass different arguments to see transcription in Punjabi ?
I use below code snippet
pipe = pipeline(model=peft_model_id, task='automatic-speech-recognition', device = 0) # change to "your-username/the-name-you-picked"
def transcribe(audio):
text = pipe(
audio,
generate_kwargs={
"task": "transcribe",
"language": "punjabi",
}
)["text"]
return text