WishLens

Inspiration

Most skincare personalization starts by analyzing a face and immediately deciding what the shopper should care about.

WishLens reverses that.

Instead of letting an algorithm silently choose the optimization target, WishLens uses Perfect Corp to create visual options on the shopper’s own face, lets the shopper choose what matters to them, and then reshapes the shopping experience around that choice.

What it does

WishLens turns skin simulation into a preference-input mechanism rather than just a visualization at the end of a recommendation flow.

1. A shopper uploads a clear face photo.
2. Perfect Skin Analysis runs live and returns structured skin-dimension scores.
3. WishLens selects eligible dimensions that are also supported by Perfect Skin Simulation.
4. Perfect generates controlled one-variable visual targets at the same fixed intensity.
5. WishLens re-analyzes each simulated image with Perfect Skin Analysis as an internal API self-consistency check.
6. The shopper chooses the visual target that matters most to them.
7. The exact same product shelf is deterministically reordered around that preference.
8. Switching targets immediately reorders the same products again.

The important interaction is simple:

The shopper does not choose a product first. They choose what matters, and the store adapts.

How we built it

WishLens is built with Next.js, React and TypeScript.

Perfect Corp is the core intelligence layer:

• Perfect File API handles image upload.
• Perfect Skin Analysis v2.1 provides live structured skin-dimension scores.
• Perfect Skin Simulation creates isolated visual target options.
• Each simulated result is uploaded and analyzed again to measure internal intervention consistency.

The simulation stage uses a fixed intensity of 0.7, and each request changes only one supported skin dimension at a time.

The retail layer deliberately uses no LLM. Products have curated concern metadata, and ranking is deterministic and explainable. Selecting a different target changes the ranking of the same nine products rather than replacing them with a new hard-coded list.

Why Perfect Corp is central

Perfect is not an add-on after the recommendation.

Perfect helps determine which visual dimensions can be compared, generates the controlled visual alternatives, and measures the resulting simulated images.

Without Perfect, the core WishLens interaction — choosing a shopping preference visually on your own face — disappears.

Challenges we ran into

Perfect has strict image requirements, so we had to validate face size, image resolution and upload behavior carefully.

The APIs are asynchronous, so the live flow also required task creation, polling, timeout handling and honest failure states.

We also wanted to avoid overstating what simulation means. WishLens does not present these images as medical predictions or guaranteed product outcomes.

The re-analysis step is only an internal API self-consistency check, not clinical validation.

Accomplishments that we're proud of

• Live Perfect Skin Analysis in production
• Live Perfect Skin Simulation in production
• Analysis-derived visual targets rather than fixed demo presets
• One-variable simulations at the same fixed intensity
• Live re-analysis of every simulated result
• Deterministic ranking of the same product shelf
• Instant target switching that visibly reorganizes the same products
• No mock Perfect results in the live golden path

What we learned

Personalization does not have to mean an algorithm deciding what the user should want.

Simulation can become an input mechanism.

Instead of saying:

“AI thinks this is your biggest concern.”

WishLens can ask:

“Which of these visual targets matters to you?”

That human choice becomes the preference signal that drives the store.

What's next

Next steps would include larger retailer catalogs, richer merchandising rules, consented preference history, and testing whether visual preference signals improve product discovery.

WishLens intentionally does not claim medical diagnosis, clinical validation, treatment prediction or guaranteed product efficacy.

Perfect measures. Perfect simulates. You choose. The store responds.

Built With

• ai
• nextjs
• perfect-corp
• perfect-skin-analysis
• perfect-skin-simulation
• react
• retail
• typescript
• vercel