From WordPress to Cloudflare, with AI-generated assets.

We rebuilt the Lam Institute for Hair Restoration from scratch as a static Cloudflare project — in twelve parallel design directions — and produced all-new identity-preserved hero photography plus palette-matched motion to accompany each one. Single source photograph of Dr. Lam in, twelve distinct visual languages out. Click through the tabs above.

The twelve design directions

12 variants · 9 pages each · stack-balanced 4 / 4 / 4

Three experimental variants (A3, B5, C2) were cut after QA. Their hero compositions left empty colour blocks instead of placing the subject in frame — a finding worth keeping, not a result worth shipping. Files preserved at sites/hairtx/variants/{A3,B5,C2} for v2 review.

Identity-preserved hero photography, twelve ways.

We fed a single source photograph of Dr. Lam to Gemini 2.5 Flash Image (Nano Banana) along with twelve detailed editorial prompts. The model held his exact face — hairline, glasses, smile — while completely changing palette, lighting, mood, wardrobe styling, and composition to match each design variant's brand direction.

Source reference

Original Dr. Lam reference photo · 3,400 × 2,400 px studio photograph · single light source · taken in clinic

One photo in. Twelve editorial heroes out →

Pipeline Model: gemini-2.5-flash-image · Input: 1 reference photo + 12 prompts · Output: 12 variant heroes at 16:9 · Cost: ≈$0.04 per image (~$0.50 total) · Turnaround: ~6 minutes start to finish · Identity check: face, hairline, glasses, smile preserved across every output.

The full AI video pipeline.

A side-by-side comparison of cloned-voice vs native-voice rendering on a real Lam dental FAQ script — plus the full V1 deliverables, V2 tool options, costs, and pipeline diagram. Embedded below; open in a new tab for the standalone view.

Open standalone ↗

Palette-matched motion for every direction.