How to get subtitles with timestamps when invoking transcribe() in python?

[Bearmancer]

I have imported whisper and this is my code:

def whisper_logic(file: Path, model: str, language: str):
    if file.suffix.lower() not in FILE_EXTENSIONS:
        return

    subtitle_file = file.with_suffix('.srt')

    if subtitle_file.exists():
        print(f"Subtitle for {file.stem} already exists. Skipping...")
        return

    print(f"Now transcribing: {file.name}")

    model = whisper.load_model(model)
    
    model.transcribe(str(file), fp16=False, verbose=False, language=language)

I want the output to be the same as whisper video.mp4 would be, i.e. srt (and other) file with timestamps.

[glangford]

I want the output to be the same as whisper video.mp4 would be, i.e. srt (and other) file with timestamps.

Not sure what whisper video.mp4 is, but I think you are looking for something like this:

from whisper.utils import get_writer
# transcribe with word timestamps
result = model.transcribe(str(file), fp16=False, verbose=False, language=language, word_timestamps=True)
# Get a writer and configure it with your preferred subtitle formatting
writer = get_writer("srt", ".")
writer(result, str(file), {"max_line_width":55, "max_line_count":2, "highlight_words":False} )

[Bearmancer]

Whisper video.mp4 is what I would write in the terminal to get it transcribed, yes? That's what I want Get Outlook for Android<https://aka.ms/AAb9ysg>

…

________________________________ From: Glenn Langford ***@***.***> Sent: Wednesday, January 29, 2025 7:28:13 AM To: openai/whisper ***@***.***> Cc: Bearmancer ***@***.***>; Author ***@***.***> Subject: Re: [openai/whisper] How to get subtitles with timestamps when invoking transcribe() in python? (Discussion #2513) I want the output to be the same as whisper video.mp4 would be, i.e. srt (and other) file with timestamps. Not sure what whisper video.mp4 is, but I think you are looking for something like this: from whisper.utils import get_writer # transcribe with word timestamps model.transcribe(str(file), fp16=False, verbose=False, language=language, word_timestamps=True) # Get a writer and configure it with your preferred subtitle formatting writer = get_writer("srt", ".") writer(result, str(file), {"max_line_width":55, "max_line_count":2, "highlight_words":False} ) — Reply to this email directly, view it on GitHub<#2513 (comment)>, or unsubscribe<https://github.com/notifications/unsubscribe-auth/AQWHKKTUISYGUXSPZEHRWE32NAYTLAVCNFSM6AAAAABWBU72JOVHI2DSMVQWIX3LMV43URDJONRXK43TNFXW4Q3PNVWWK3TUHMYTCOJZGA3DOMA>. You are receiving this because you authored the thread.Message ID: ***@***.***>

[glangford]

Whisper video.mp4 is what I would write in the terminal to get it transcribed, yes?

Ah, now I get it. :) Sorry for the misunderstanding. The default option from the command line is to produce all outputs.

This is how whisper writes out everything:

whisper/whisper/utils.py

Line 307 in 517a43e

if output_format == "all":

So try something like this

from whisper.utils import get_writer
result = model.transcribe(str(file), fp16=False, verbose=False, language=language, word_timestamps=True)
all_writers = get_writer("all", ".")
all_writers(result, str(file), {"max_line_width":55, "max_line_count":2, "highlight_words":False} )