The original oblique movie and Sweep27's observations remain the source of motion constraints.
A rejected camera edit must not contribute a new paw, ankle, hip, pelvis or chest angle.
A matching motion rhythm with an unchanged camera adds no observation of the hidden side.
A successful viewpoint change with a new face, new tools or a different action describes a new
performance, not another camera of this performance.
- Preserve the exposed support silhouette and sole height while the working arm and broom travel.
- Preserve two work/recovery phases in the studied interval and the more open recovery pose.
- Preserve one hand holding one unchanged broom at a consistent shaft location.
- Use the whole visible character, including the accepted head design, limb proportions and tail.
- Treat changing gray-body coverage over black paws as occlusion unless separate evidence shows
actual foot deformation, rocking or rotation.
The primary view does not uniquely separate foot yaw, ankle accommodation, hip rotation,
pelvis yaw, chest yaw or head counter-rotation. A narrowing belly patch or differential back
edge displacement can be explained by occlusion and common-body inclination as well as by
articulation. No video-edit output is a force, balance, friction or penetration measurement.
Consequently, do not author new twisting feet merely because the hidden side is uncertain.
Do not adopt a mandatory pelvis/chest separation solely from a generated shape change. Do not
mechanically fit image-feature centroids as joint targets.
Retain the seven-pose experiment specified in Sweep27: originalframes0,42,61,75,90,106,142.
Compare a common body frame with a lower-body/chest articulation candidate under the same
supported-foot and hand/tool tasks. Author semantic controls for the working inclination,
recovery opening and shoulder/neck response. Keep broad, smooth torso/hip transitions.
Then ablate chest separation and leg response individually to identify which control improves
actual source behavior. Add a foot-rotation study only if a remaining observed mismatch needs it.
Judge every candidate with the same editable geometry from front, source-oblique, side and rear
cameras, including isolated shaded hip/ankle surfaces. Inspect attachment, tool continuity,
contact, silhouette and source performance before rendering a complete loop. These are proposed
experiments; no new native controls or geometry were implemented in Sweep28.
The generated movies and exact image guides remain review material. No accepted SSOT pointer,
release, native baseline or motion registry is replaced by this study.