Six body controls.
Still no fitted wardrobe.

V29 reproduced the V28 failures and compared six smaller regions. V30 made hard-row feasibility reproducible and reduced the artificial isolated bust snap, but its joint field crossed the body. Neither checkpoint justified a whole-front solve.

A garment can only be fitted to a body state that has been chosen and preserved. Continuing to optimize clothing around provisional proportions would make every clearance, landmark and strain result temporary. V31 changes the order of work: establish reversible body controls first, then rebuild body-dependent garment evidence after a preferred target is reviewed.

Both source bodies contain 19,158 vertices and 163-bone rigs, with 294 vertex groups and their existing shape keys, UVs and skin weights retained. Evaluated heights are 1.617863 m for A and 1.583931 m for B. Their inherited macro states already differ, so V31 records separate source hashes and never treats one body as the other’s default.

Six new relative shape keys sit above those sources. Zero returns the evaluated body exactly to baseline. Fresh native copies reopened with working deformation, a measured reset error of 0 mm and protected face vertices unchanged. This preserves the rig and topology; it does not certify the anatomy or every possible control combination.

The V31 schema exposes breast fullness, glute volume, thigh thickness, calf development, hip width and waist shape. Each has a reversible 0–1 draft range, baseline and draft presets, plus editable custom JSON. The private Character Studio provides controls to switch A or B, compare original and current states, change four views, show the reference, toggle clothing and wireframe, recalculate the preview and save a full studio copy.

Background operator checks exercised body updates, preset save/load, source validation and exact reset. An update plus evaluated readback measured 31.90–62.20 ms across 20 local cases including dependency-graph work. That is not a foreground frame-rate or mobile benchmark. At the V31 checkpoint, physical mouse interaction and Blender’s Undo/Redo hotkeys were not tested.

A safe derivative of each earlier skirt follows the six body parameters through corresponding keys. That proves synchronization. It does not prove fit. At the draft preset, source-relative evaluated triangle strain spans 0.1413×–13.2095× for A and 0.0435×–6.5303× for B.

Those ranges include thin solidified boundary triangles and are diagnostic values, not fabric strain qualification. The standing static tests found zero body pairs, but excessive ease and invalid strain keep both results failed. A zero collision count cannot override the other measurements, just as it could not in V28.

Six synthetic diagnostic pose states per body remained finite while preserving each original rig. Two saved sitting renders expose the more useful truth: visible upper-thigh compression, creasing and glute deformation need corrective sculpting and skinning review.

No production walk or sit action was authored. Body self contact, full joint-volume quality, facial animation, runtime behavior and mobile performance are not certified.

The next meaningful decision is visual: inspect A and B in the private studio, adjust the proportions and save a preferred preset for review. After that selection, the lower wardrobe needs region-aware correspondence and material-control-surface strain evaluation. Body landmarks, normals, collision geometry and every body-dependent garment result must be reconstructed for the chosen state.

The upper prototype remains blocked because the first garment-fit gate failed, and full-blouse fitting remains paused until body approval. V31 does not make either receptionist complete. It makes the dependency visible and gives the next review an editable, reproducible foundation.

Later V32 usability caution · October 22: the reported approximately 60-second owner freeze was not reproduced, and its root cause remains unresolved. V32 removed synchronous detailed validation from the editor, excluded the inactive character collection and defaulted to a body-only preview. Limited foreground work observed a numeric edit, character switch, reset and Undo; it did not qualify continuous dragging, Redo, a cold launcher run or the final byte revision. Sustained foreground editing is the next gate before body selection or blouse work resumes.