Archive footage beside its doodle cartoon. The full shot is the motion lane: a rig in-betweened with the footage's physics. The ten drawn subjects are the look lane: traced keys, played held or in-betweened by matching strokes from key to key. Both are measured on the same scale on the right.
The giraffe method, done by hand for all ten: the chosen doodle itself, every stroke kept, moved by a few bones placed on the footage key by key, background drawn once. Then the same tests the look lane ran, plus a blind critic who saw only an eight-frame strip of the cartoon and had to name what it was.
| Subject | M2 look-lane keys | M2 puppet | T3 strokes carry | B1 background holds | J4 same doodle | J1 critic named it | M5 proportions | M3 footage fit |
|---|
Look lane / puppet where both ran. M2 is the percentile among real doodles of the category (bar 30). M3 uses a GrabCut silhouette, not SAM, and is weak. The critic strip skips frames, so a "broken" motion verdict is partly the strip's spacing; playback review is still open.