Is it possible to set initial_prompt and condition_on_previous_text with a whisper_pipeline?

Is it possible to set initial_prompt and condition_on_previous_text with a whisper_pipeline?

i know this can work:

whisper_pipeline = pipeline(“automatic-speech-recognition”, model=model_name, torch_dtype=torch_type, device_map=“auto”, model_kwargs=model_args)

whisper_pipeline(speech_file, generate_kwargs = {“task”:“transcribe”, “language”:“<|fr|>”} )

but is it possible to set initial_prompt and condition_on_previous_text parameters?

2 Likes

Hi, is there any development on this?

Same here, any update?

yup check out the github topic on this

Could you send a link please?