# Static Ad Briefs V3 — Sleep + Recovery (latest)

V3 supersedes static-ad-briefs.md. Reflects current canonical_prompt.py + tier-2 reference chain learnings.

## Per-frame production process (v3)

1. Read `products/sleep-recovery-magnesium/canonical-prompt.yaml`
2. Define scene_type from SCENE_SCALES (hero / editorial / lifestyle / held / wild / amazon-spec)
3. Build scene description (focus on composition + lighting, NOT pack technical specs)
4. Call `canonical_prompt.build_prompt(...)` to compose full prompt
5. Fire Nano Banana with multi-ref image_input chain (v0 + tier-2)
6. Verify output against unified pack-rendering law (aspect + scale + perspective + front-of-pack)
7. If drift: curate scene_type or refire at smaller pack scale

See `static-ad-briefs.md` for the actual per-Lane brief content.