// a real episode, published
One episode, opened all the way up.
The annotated session, opened up: the raw clip, the product structuring it, and the layer-by-layer anatomy of the episode it becomes. No demo data, no mockups.
Everything on this page is real — the raw input clip and the annotation session on it. The packaged episode JSON publishes next.
raw clip · input
the session · in review
The layers
Each panel below is one layer of this episode. Schematics are coded wireframes — labels and values come from the published JSON.
L0 · RAW FRAME
The egocentric frame as captured — the part everyone already has.
L1 · CONSENT
A per-episode consent packet: who appears, what they agreed to, what rights attach.
L2 · OBJECT TRACKS
Every object detected and tracked frame-to-frame with a stable identity.
L3 · HANDS
Both hands detected and tracked as 21-keypoint poses, grasp by grasp.
L4 · TEXT CAPTIONS
An action-aligned natural-language description, anchored to the frames.
Want 5 of these on your task?
Request a sample episode→