Batch transcribe audio and video files using OpenAI Whisper

transcribe-audio-whisper.py

import os
import math
import subprocess
import openai
import pysrt
from datetime import timedelta

API_KEY = "api-key"
INPUT_DIR = "input_media"
CHUNK_DIR = "audio_chunks"
OUTPUT_DIR = "output"
CHUNK_DURATION_SEC = 300

client = openai.OpenAI(api_key=API_KEY)
os.makedirs(CHUNK_DIR, exist_ok=True)
os.makedirs(OUTPUT_DIR, exist_ok=True)

def list_media_files(directory):
    return [
        os.path.join(directory, f)
        for f in os.listdir(directory)
        if f.lower().endswith(('.mp4', '.mp3'))
    ]

def get_duration_seconds(input_path):
    cmd = [
        "ffprobe", "-v", "error",
        "-show_entries", "format=duration",
        "-of", "default=noprint_wrappers=1:nokey=1",
        input_path
    ]
    return float(subprocess.check_output(cmd).decode().strip())

def split_media(input_path, output_dir, chunk_duration):
    total_duration = get_duration_seconds(input_path)
    total_chunks = math.ceil(total_duration / chunk_duration)
    base = os.path.splitext(os.path.basename(input_path))[0]
    chunk_paths = []

    for i in range(total_chunks):
        start_time = i * chunk_duration
        output_file = os.path.join(output_dir, f"{base}_chunk_{i:03}.mp3")
        cmd = [
            "ffmpeg", "-y", "-i", input_path, "-ss", str(start_time),
            "-t", str(chunk_duration),
            "-vn", "-ar", "16000", "-ac", "1", "-b:a", "64k",
            output_file
        ]
        subprocess.run(cmd, stdout=subprocess.DEVNULL, stderr=subprocess.DEVNULL)
        chunk_paths.append(output_file)

    return chunk_paths

def transcribe_chunk(path):
    with open(path, "rb") as f:
        return client.audio.transcriptions.create(
            model="whisper-1",
            file=f,
            response_format="srt"
        )

def shift_srt(srt_data, shift_seconds):
    subs = pysrt.from_string(srt_data)
    for sub in subs:
        sub.shift(seconds=shift_seconds)
    return subs

def process_file(filepath):
    base_name = os.path.splitext(os.path.basename(filepath))[0]
    print(f"Processing: {base_name}")

    chunk_paths = split_media(filepath, CHUNK_DIR, CHUNK_DURATION_SEC)
    all_subs = pysrt.SubRipFile()

    for i, chunk_path in enumerate(chunk_paths):
        print(f"  Transcribing chunk {i + 1}/{len(chunk_paths)}")
        try:
            srt_text = transcribe_chunk(chunk_path)
            shifted = shift_srt(srt_text, i * CHUNK_DURATION_SEC)
            all_subs.extend(shifted)
        except Exception as e:
            print(f"  Error on chunk {i}: {e}")

    final_srt_path = os.path.join(OUTPUT_DIR, base_name + ".srt")
    all_subs.save(final_srt_path, encoding="utf-8")
    print(f"Saved: {final_srt_path}\n")

def main():
    media_files = list_media_files(INPUT_DIR)
    print(f"Found {len(media_files)} media files to process\n")
    for f in media_files:
        process_file(f)

if __name__ == "__main__":
    main()