Inspiration
Video has been part of my work for years, and the most tedious part of editing is not editing: it is logging. An interview shoot produces 4–8 hours of footage, and someone has to watch it all, note timecodes and type the key lines before the story can be built. Cloud transcription is often not an option, because unreleased footage is private or under NDA.
What it does
Footage Paper Edit is a UGOS Skill that turns a folder of raw footage on the NAS into a timecoded paper-edit log:
- each clip's duration, resolution, orientation and audio presence
- the spoken lines with start and end times
- optional clip tags: dialogue, no-audio, vertical, 4k
Then the editor can ask "where does Maria mention the rain?" and get the clip name and the minute.
How we built it
It is a pure orchestration Skill built with the official UGOS Skill Creation Tool (uliya-skill-kit 0.7.3). It delegates only to the NAS experts: file-agent finds the clips and reads media info, and knowledge-agent transcribes, saves the log and answers follow-up questions. Footage is read-only, saving and tagging go through NAS approval, and tagging is idempotent. The package passes the kit's static check with 0 errors and 0 warnings.
Challenges
Designing for real-world mess: silent clips, corrupted files, partial transcription failures and very large folders. Every step reports empty or partial results honestly and never invents data.
What we learned
A good Skill is less about prompts and more about workflow design: clear scopes, approvals and failure states.
What's next
Testing on a real AI NAS: confirming that transcription accepts video files directly and returns segment timestamps, then tuning the log on real shoots.
Built With
- markdown
- ug-filemgr
- ug-filesystem
- ug-knowledge
- ugos
- uliya-skill-kit