Inspiration
Scene Shelf is for indie animators and small studios. My own animated feature, Lady Elise, has a finished animatic with 3,045 storyboard panels and 1,123 recorded voice lines. Every one of those has to sit in the right scene. On a small team, that is a lot of folders to keep straight, and one missing line can stall a scene without anyone noticing until the edit.
What it does
Scene Shelf sorts the raw pieces of an animated film — storyboard panels, recorded voice lines and the script — into one folder per scene on the NAS, then writes a short review sheet for each scene.
How it works
Choose a project folder. Choose the script file. Scene Shelf proposes a scene list, and you confirm it. It sorts copies, then shows a summary per scene: panel count, lines found, lines missing, and unmatched files. You fix the flagged items and run it again.
The Skill reads the script, finds the scene headings, then matches panels and audio to scenes using file names, dates and text it can read. It moves copies (never originals) into scene folders and flags gaps: a line in the script with no recording, a panel with no scene.
What makes it different
Generic AI tools organize photos by faces or dates. Nothing I have used understands that a film is built scene by scene from three kinds of files that have to line up. Scene Shelf uses the script as the map.
How we built it
Build status: Idea stage. The Skill will be newly built during the Hackathon with the UGREEN Skill Developer Package and an AI coding assistant, disclosed in SKILL.md. SKILL.md drives the workflow: parse the script into scenes, list the media, match by name patterns first, then by transcribed or extracted text where available. Copies only; originals stay untouched. It uses local storage for large media sets, so nothing leaves the user's own hardware.
Test plan
Core test 1: a 3-scene sample script with 30 panels and 20 voice files sorts correctly, with a correct missing-line report. Core test 2: running it twice gives the same result and makes no duplicate copies. Edge case: files with no scene clue in name or content land in an "Unmatched" folder, never in a wrong scene.
Known limits
Badly named files with no readable text cannot be matched. Scripts that are not in standard screenplay format may need the user to confirm scenes by hand. Large audio sets may be slow to transcribe on the device.
Built With
- ugos
- ugreen
