Lilycut is an AI video editor I’m building for recordings of hands-on projects. I want it to find the useful parts of a long recording and put together a story that shows what happened.
What I want it to do
From the same footage, it should make a 30-second transformation, a two-minute overview of the process, and a ten-minute version with more explanation, problems, and decisions. Spoken markers and commentary should help it find important moments, but it also needs to work without speech.
I want to review and adjust the shots, reuse the analysis for different edits, and render from the original files. Cost, processing time, and the amount of manual correction all matter.
Starting point
The starting requirements describe the first release and the initial test: a useful two-minute edit from a representative project recording. I’ll update that post as I build and test the editor.
The original idea stays in the Ideas backlog and is also part of this project’s history.
If I add a command-line interface, its short command will be lily.
Follow the Lilycut project feed for requirements, experiments, and later updates.