transcribe.py

import argparse
from pathlib import Path
import whisper
from whisper.utils import WriteSRT

modelSize = "large"

if __name__ == "__main__":
    parser = argparse.ArgumentParser(description="Synchronize SRT timestamps over an existing accurate transcription.")
    parser.add_argument('pathMp3', type=str, help="Path to the SRT file with good timestamps")
    parser.add_argument('modelSize', type=str, help="Path to the TXT file with good text", nargs='?')
    args = parser.parse_args()
        
    if args.modelSize != None:
        modelSize = args.modelSize
        
    print("Loading Whisper model: "+modelSize+" ...")
    model = whisper.load_model(modelSize)
    
    print("Transcribing: "+args.pathMp3+" ...")
    result = model.transcribe(args.pathMp3)
    
    print(result["text"])
    
    print("Saving: "+args.pathMp3+".srt ...")
    p = Path(args.pathMp3)
    writer = WriteSRT(p.parent)
    writer(result, args.pathMp3)