Quick and dirty script to find files with missing subtitles and transcribe them with whisper.cpp -- requires ffmpeg and whisper.cpp. Run like: python3 transcribe-missing.py path/to/media/files

make-subtitles.sh

#!/bin/bash
set -euxo pipefail

echo "Converting audio"
rm -f temp.wav
ffmpeg -i "$1" -ar 16000 -ac 1 -c:a pcm_s16le temp.wav
echo "Transcribing"
./main -m models/ggml-base.en.bin -f ./temp.wav --output-srt -t 8 -ml 42
mv temp.wav.srt "${1%.*}.en.srt"
# optional: rewrite the original file to include subtitles
#echo "Adding to video file"
#ffmpeg -i "$1" -i temp.wav.srt -c copy -metadata:s:s:0 language=eng "${1%.*}.subtitled.mkv"

transcribe-missing.py

#!/usr/bin/env python3

import json
import os.path
import os
import subprocess as sp
import sys
import time


def has_subtitles(fname) -> bool:
    base, _ = os.path.splitext(fname)
    if os.path.exists(base + '.srt') or \
       os.path.exists(base + '.en.srt') or \
       os.path.exists(base + '.english.srt') or \
       os.path.exists(base + '.stt') or \
       os.path.exists(base + '.en.stt') or \
       os.path.exists(base + '.english.stt') or \
       os.path.exists(base + '.sub'):
        return True
    out = sp.check_output(['ffprobe', '-v', 'quiet', '-print_format', 'json', '-show_format', '-show_streams', fname], stderr=sp.STDOUT)
    for stream in json.loads(out)['streams']:
        if stream['codec_type'] == 'subtitle':
            return True
    return False


check_endings = {'.mkv', '.mpg', '.avi', '.mp4', '.m4v', '.mov'}
for dirpath, dirnames, fnames in os.walk(sys.argv[1]):
    for name in fnames:
        fname = os.path.join(dirpath, name)
        _, ext = os.path.splitext(fname)
        if ext in check_endings:
            if not has_subtitles(fname):
                print(fname)
                sp.check_call(['./make-subtitles.sh', fname])