JimmySadek/video-fetcher-to-markdown

What it solves

Video Fetcher to Markdown converts video links from various platforms into structured, archival Markdown notes. It solves the problem of raw transcript extractors that lack essential context, such as creator metadata, descriptions, and chapters, by consolidating all this information into a single, portable file suitable for knowledge bases like Obsidian or Logseq.

How it works

The tool operates differently based on the source:

• YouTube: It reads captions directly from YouTube's interface without requiring an API key.
• Other Platforms: For sites like Instagram, TikTok, X, Vimeo, and Facebook, it uses yt-dlp to download the media, ffmpeg for audio conversion, and a local instance of Whisper (via mlx-whisper or openai-whisper) to transcribe the audio on the user's machine.
• Visuals: For short videos, it generates a "contact sheet" of frames to capture on-screen text and visual cues.
• Output: It generates Markdown files with YAML frontmatter for queryability (e.g., via Dataview) and supports multiple export formats including JSON, SRT, and WebVTT.

Who it’s for

Researchers, students, and knowledge workers who use personal knowledge management (PKM) systems like Obsidian or Logseq and want to create a permanent, searchable archive of video-based learning materials.

Highlights

• No API Keys: Works without hosted services or API keys.
• Multi-platform Support: Supports YouTube and any site compatible with yt-dlp.
• Local Transcription: Uses Whisper for local, private transcription of non-YouTube videos.
• Archival Quality: Captures title, channel, duration, upload date, description, and chapters.
• Language Flexibility: Supports ordered language preferences, automatic selection, and YouTube machine translation.
• Visual Context: Includes a contact sheet of frames for short videos to capture on-screen information.