Caption

Caption provides direct transcription for episodes that will remain on YouTube rather than migrating to Mux. Using yt-dlp for audio download and the Gemini File API for transcription, Caption produces four outputs per episode: full transcript text, AI-generated summary, extracted key points, and Complexity Trap detection. This path is used for historical episodes where Mux migration is not planned. Caption works in batch mode with configurable sizes and supports show-specific targeting. After transcription completes, Vault syncs the transcript into the content database and Forge scans it for article opportunities.