How to transcribe and translate at the same time to create two separate files?

[eric-bks]

In Python, preferably.

I have extensive files and don't want to wait twice as long to run the same command. How do I get Korean and English SRT files from my function?

Here's what I usually use:

def transcribe():
    model = whisper.load_model("large", device = 'cuda')
    audio = whisper.load_audio(f'{output_name}.mp4')
    global result
    result = model.transcribe(audio, verbose = True, language = 'Korean', task = 'translate')

I've tried this: model.transcribe(audio, verbose = True, language = 'Korean', task = 'transcribe' + 'translate'), but that doesn't work, and I imagine it would still take as much time as running it separately.

[phineas-pta]

u can open 2 different python process, 1 for transcribe 1 for translate

[eric-bks]

I still share resources, so they would both run half as fast, and this wouldn't save time.

Doesn't whisper have to transcribe first before it can translate? So I imagine that the transcription is already created when performing the translation. I should be able to output both from one operation.

[phineas-pta]

Doesn't whisper have to transcribe first before it can translate?

no, whisper was trained to output directly translation from audio, u can read the paper

[spiteria]

@phineas-pta Where can I find the paper you are referring to?

[phineas-pta]

[View-my-Git-Lab-krafi]

any update in one year