kartikeysepta/youtube-transcript-scraper

This is a Python command-line tool that takes a YouTube video link and gives you back a structured JSON file containing the video's metadata and a full timestamped transcript. You run a single command with the video URL and it handles the rest: downloading the audio, converting it to MP3, pulling the channel and video details, and transcribing the spoken words with timestamps. Transcription can happen in two ways. The default method runs a speech-to-text model called Faster Whisper directly on your machine. This works without any external account or API key, and it can use a graphics card to speed up processing if you have one available. The second option sends the audio to Google Gemini, a cloud AI service, which requires a free Google AI Studio API key that you add directly to the script. The output JSON includes fields like title, channel name, view count, like count, upload date, tags, categories, and the full transcript with time markers so you can find specific moments. When you point the tool at an existing JSON file, it appends the new result rather than overwriting, making it easy to build a growing dataset across multiple videos. The README lists several practical purposes: building a searchable archive of video transcripts, extracting metadata for SEO or content analysis, creating datasets for AI search systems, and converting long lectures or interviews into text you can work with in other tools. Setup requires Python, FFmpeg installed on your computer, and cloning the repository. The API key for Gemini mode is currently hardcoded in the script rather than read from an environment variable, which the contributing notes flag as a known improvement to make. The project has no license file yet, so the terms for reuse are unclear until one is added.