Generate subtitles/captions for MP4 file using OpenAI Whisper

speech_to_text.py

import os
import whisper
from pathlib import Path
from concurrent.futures import ProcessPoolExecutor

#Requirements:
#pip install -U openai-whisper
#winget install ffmpeg
#copy ffmpeg.exe from this folder
#C:\Users\foo\AppData\Local\Microsoft\WinGet\Packages\Gyan.FFmpeg_Microsoft.Winget.Source_8wekyb3d8bbwe\ffmpeg-6.0-full_build\bin\ffmpeg.exe

#May also need:
#pip install torch
#pip install more-itertools numba numpy tiktoken tqdm

#To update Whisper:
#cd C:\Python39\Scripts
#./pip install --upgrade --no-deps --force-reinstall git+https://github.com/openai/whisper.git

def transcribe_file(input_file, model):

    input_filename = os.path.basename(input_file)
    output_folder = os.path.dirname(input_file)

    try:
        print("Processing: " + input_filename)
        
        txt_output_file = Path(input_file).with_suffix('.txt')   
        srt_output_file = Path(input_file).with_suffix('.srt')
        
        if os.path.isfile(txt_output_file):
            print('Output already exists: ' + os.path.basename(txt_output_file))
            return
        
        result = model.transcribe(input_file)
        
#        with open(txt_output_file, 'w') as f:
#            f.write(result["text"])

        srt_writer = whisper.utils.get_writer("srt", output_folder)
        srt_writer(result, srt_output_file)
        txt_writer = whisper.utils.get_writer("txt", output_folder)
        txt_writer(result, txt_output_file)
            
    except Exception as e:
        print("[" + input_filename + f"] An error occurred: {type(e).__name__} - {str(e)}")

def main():
    input_folder_str = r"H:\Videos"
    
    input_folder = Path(input_folder_str)
    
    input_extensions = ['.mp4', '.avi', '.mkv', '.wav']
    input_files = [file for ext in input_extensions for file in input_folder.glob(f'**/*{ext}')]

    #model = whisper.load_model("large")	#takes a long time to process using this model
    model = whisper.load_model("base")
    

    # Use a ProcessPoolExecutor to run the transcriptions in parallel
    with ProcessPoolExecutor(max_workers=4) as executor:   
        for input_file in input_files:
            #Seems to crash if run in parallel. Not sure exactly why, but I think github discussion said the module wasn't thread safe
            #executor.submit(transcribe_file, input_file, model)
            transcribe_file(str(input_file), model)

if __name__ == "__main__":
    main()