# V97 RC11 — Pose Sprite Inventory & Curation

## Purpose

This release does **not** continue the unique AI image batch workflow. It inventories the already-generated athlete figures, extracts reusable pose candidates, and prepares a pose-sprite direction for future motion fallback.

Runtime remains unchanged: pass view and library still load regular exercise figures via `assets/img/exercises/figures/<exerciseKey>.webp`.

## Added assets

- `assets/img/exercises/pose-sprites/candidates/alpha/`  
  Auto-extracted pose candidates from images that already had usable alpha.

- `assets/img/exercises/pose-sprites/candidates/keyed/`  
  Auto-extracted pose candidates from full-canvas / greenscreen-like sources. These are useful as references but need cleanup before active use.

- `assets/data/image-admin/pose-sprite-curation.json`  
  Curated grouping, best-pick notes and prototype motion sequences.

- `assets/data/image-admin/pose-extract-inventory-alpha.json`  
  Raw extraction inventory from alpha-ready images.

- `assets/data/image-admin/pose-extract-inventory-keyed.json`  
  Raw extraction inventory from keyed/full-alpha images.

## Contact sheets

- `docs/pose_sprites/V97_RC11_FIGURES_CROPPED_CONTACT_SHEET.png`  
  Overview of current full exercise figure assets.

- `docs/pose_sprites/V97_RC11_POSE_CANDIDATES_ALPHA_CONTACT_SHEET.png`  
  Candidate pose crops from already-transparent assets.

- `docs/pose_sprites/V97_RC11_POSE_CANDIDATES_KEYED_CONTACT_SHEET.png`  
  Candidate pose crops from greenscreen/white/full-canvas sources.

## First curation notes

### Best generic squat / load pose

Primary pick: `broad-jump__pose05.webp`

Reason: good generic unloaded squat/load/landing posture without dumbbell or box.

### Best loaded squat pose

Primary pick: `goblet-squat__pose02.webp`

Reason: strong squat mechanics, but includes a dumbbell and should be treated as loaded-goblet-specific.

### Best bear crawl sequence

Primary picks: `bear-crawl__pose01.webp`, `bear-crawl__pose02.webp`

Reason: clean two-phase crawl sequence with good full-body readability.

### Best push-up / plank sequence

Primary picks: `push-up__pose01.webp`, `push-up__pose02.webp`, `push-up__pose04.webp`

Reason: clean top-bottom-top movement sequence.

### Row sequence

Primary picks: `single-arm-row__pose01.webp`, `single-arm-row__pose02.webp`

Status: motion is useful, but source was greenscreen/full-canvas and needs edge cleanup before active runtime use.

## Recommended next step

Build a small **Motion Sprite Prototype** without committing to full runtime replacement:

1. Use 4–6 curated sequences:
   - `bear-crawl`
   - `push-up`
   - `goblet-squat`
   - `linear-pogos` after cleanup
   - `sit-up`
   - `side-plank-hip-lift`
2. Render them inside the existing 16:9 media container.
3. Use them only when no video and no approved final figure exists.
4. Keep current exercise figures as the active visual path until the pose-sprite renderer is explicitly accepted.

## Important

Do not blindly follow earlier image-batch handoff notes that suggest generating more unique exercise images. This RC intentionally pivots to extracting value from existing figures and grouping reusable poses.
