import re from googleapiclient.discovery import build from googleapiclient.errors import HttpError from youtube_transcript_api import YouTubeTranscriptApi from dotenv import load_dotenv import os import json import isodate import argparse def get_video_id(url): # Extract video ID from URL pattern = r'(?:https?:\/\/)?(?:www\.)?(?:youtube\.com\/(?:[^\/\n\s]+\/\S+\/|(?:v|e(?:mbed)?)\/|\S*?[?&]v=)|youtu\.be\/)([a-zA-Z0-9_-]{11})' match = re.search(pattern, url) return match.group(1) if match else None def main_function(url, options): # Load environment variables from .env file load_dotenv(os.path.expanduser('~/.config/fabric/.env')) # Get YouTube API key from environment variable api_key = os.getenv('YOUTUBE_API_KEY') if not api_key: print("Error: YOUTUBE_API_KEY not found in ~/.config/fabric/.env") return # Extract video ID from URL video_id = get_video_id(url) if not video_id: print("Invalid YouTube URL") return try: # Initialize the YouTube API client youtube = build('youtube', 'v3', developerKey=api_key) # Get video details video_response = youtube.videos().list( id=video_id, part='contentDetails' ).execute() # Extract video duration and convert to minutes duration_iso = video_response['items'][0]['contentDetails']['duration'] duration_seconds = isodate.parse_duration(duration_iso).total_seconds() duration_minutes = round(duration_seconds / 60) # Get video transcript try: transcript_list = YouTubeTranscriptApi.get_transcript(video_id) transcript_text = ' '.join([item['text'] for item in transcript_list]) transcript_text = transcript_text.replace('\n', ' ') except Exception as e: transcript_text = "Transcript not available." # Output based on options if options.duration: print(duration_minutes) elif options.transcript: print(transcript_text) else: # Create JSON object output = { "transcript": transcript_text, "duration": duration_minutes } # Print JSON object print(json.dumps(output)) except HttpError as e: print("Error: Failed to access YouTube API. Please check your YOUTUBE_API_KEY and ensure it is valid.") def main(): parser = argparse.ArgumentParser( description='vm (video meta) extracts metadata about a video, such as the transcript and the video\'s duration. By Daniel Miessler.') parser.add_argument('url', nargs='?', help='YouTube video URL') parser.add_argument('--duration', action='store_true', help='Output only the duration') parser.add_argument('--transcript', action='store_true', help='Output only the transcript') args = parser.parse_args() if args.url: main_function(args.url, args) else: parser.print_help() if __name__ == "__main__": main()