# Wonder v51 — reviews + growth

Five new complete generated images: A pink part-line comparison, B cobalt curls, C coral temple comparison, D lilac portrait, E yellow crown comparison. Photography, people, bottle, lettering, review panel and price badge were generated together. The final PNGs are exact copies of the selected image-generation outputs; only the review previews were resized and compressed. There was no bottle cutout, pasted product, Remotion text or post-generation artwork repair.

The prior GetHooked study remains the creative basis: 74 records, 73 visually inspected images, 28 advertisers, with 17 examples retained on this board. This turn did not collect another GetHooked sample. Obvi O06/H04/O28 informed the heavy two-color typography, supporting proof blocks and distinct offer shapes. Competitor presence, archive duration and tool labels do not establish conversion performance.

The operator preferred v50. The missing layer was trust: an ingredient strip does not do the job of a review stack. This version replaces that strip with five star symbols, 4.8/5, 1,100+ reviews, Wholesale Botanics and on Trustpilot. The $14 first-bottle offer retains its subscription qualifier. No time-to-result or invented testimonial has been added.

Review source: https://www.trustpilot.com/review/wholesalebotanics.com — checked 22 September 2026: 4.8/5 and 1,125 reviews. The displayed 1,100+ count is conservative. These are company-level reviews, not Wonder-specific hair-regrowth evidence or reviews by the depicted models.

A, C and E are generated before/after concept mockups of fictional adults. B and D also use generated models. No image represents a real customer result. The surrounding board identifies this provenance; the artwork itself has no internal-review annotation, as requested.

Initial A, C and E images were rejected because supporting content fell below the intended clearance. Their complete replacement generations moved proof and price above the comparison photographs. Initial D was rejected for a duplicated word in the attribution; the complete replacement corrected it. All rejected outputs remain in the generation directory and are recorded in lineage.json, but are not on the board.

## Placement and claim status

The selected masters are nominal 4:5 (1122×1402 pixels; approximately 0.036% ratio deviation from exact 4:5). The original bytes are preserved. These are feed concepts only. Critical copy, review attribution, offer qualifier, faces and bottle labels are above the bottom 20% house reserve. The board has both an illustrative feed frame and a separate labeled bottom-20% stress overlay; neither is an official Meta placement preview.

Actual Ads Manager placement previews and approval were not obtained. Official Meta guide/help fetches required login or returned rate-limit errors. Strong regrowth promises and generated comparison concepts remain proposed review copy, not claims established by the company rating. Customer-supplied result photography and the team's claim review are separate from this layout review. No campaign, ad set, creative object or Shopify object was changed.

## Reusable method

See method.md on the board for the seven-step workflow and a copyable prompt template. The change was more specific whole-ad direction and iterative rendered-image review, not model retraining or permanent learning. The depth/occlusion seen in these whole-image outputs was generated by the model; it was not a later hand-cut mask.