# framedex

framedex is a Claude Code skill and command-line tool that turns a scattered video and photo archive into a portable, plain-text knowledge base, writing a Markdown sidecar next to each file with transcript, location and scene description.

framedex builds a queryable, plain-text knowledge base out of a video and photo archive spread across drives and years. For each video clip, it writes a .description.md sidecar containing GPS location and place name, a speaker-diarized multilingual transcript with an English translation when needed, face detection, and an AI vision scene description with a keep, review or cull rating; still photos get the same treatment minus audio, plus a camera and exposure block read from EXIF. Sidecars live next to the originals, which are never modified, and the process is local-first, non-destructive and resumable.

The per-clip pipeline chains ffprobe and exiftool for metadata, Nominatim for reverse geocoding, ffmpeg for representative frame and audio extraction, WhisperX for transcription, translation and speaker diarization, insightface for face embeddings, and a vision model for the final structured scene description. framedex installs as a Claude Code skill providing the fdx command-line tool, and an accompanying fdx-mcp server can expose an indexed archive to Claude Desktop, LM Studio or another MCP host.

Website: https://github.com/Simbastack-hq/framedex