K-Glow Match

Inspiration

Choosing skincare online can feel overwhelming. We wanted to turn a skin report into a small, understandable next step: a curated Korean beauty routine, followed by a chance to explore lip and blush colors on the same photo.

What it does

K-Glow Match is a mobile-first beauty prototype in English, Korean and Spanish. With explicit permission, a user uploads a selfie for YouCam Skin AI analysis. The app shows five returned concern scores — moisture, dark spots, dark circles, wrinkles and texture — and skin age when the API supplies it. Missing values remain unavailable.

The report leads to concern-matched recommendations from a built-in catalog of eight real K-beauty products. An optional age-range selector refines the product ordering without changing the analysis. Users can then choose lipstick and blush colors, explicitly render a YouCam Makeup VTO result, and compare it with the original. A demo bag supports quantities and totals without taking payments or creating orders.

An explicitly labelled sample mode lets judges explore the interface without providing a selfie or spending API units. Its illustrative color preview is separate from the real YouCam render.

How we built it

The interface uses React and Vite, deployed on Vercel. Server-side API functions handle YouCam authentication, signed file uploads, task creation and bounded polling. The integration uses both YouCam Skin Analysis REST and Makeup VTO REST. API credentials remain on the server. Signed session references expire after one hour.

A small editable product catalog drives concern matching. UI translations are stored separately and the selected language is remembered on the device. Responsive layouts use large touch targets, a focused sequence of screens and a clean beauty-brand visual style.

Challenges

We had to distinguish image rejection, expired sessions, request limits and other API failures without replacing errors with fabricated results. Mobile photo decoding also needed a fallback. A skin-age response variant used a numeric score inside the explicit skin_age entry; supporting that format fixed missing age display while keeping concern scores separate from age.

Accomplishments

Both YouCam integrations have completed hosted tests using the vendor's public sample. The owner also confirmed a live phone skin report with all five concern scores, an average score and returned skin age after the parsing fix. Automated checks cover API adapters, explicit age extraction, mobile decoding, translations and recommendation behavior.

What we learned

A useful beauty flow needs understandable feedback as much as an API connection. Explicit consent, honest missing-data states, clear separation of sample and live results, and a short report-to-recommendation path make the experience easier to evaluate.

What's next

Improve product coverage and recommendation transparency, add durable per-user API usage limits before broad distribution, and validate the experience across more physical devices. Product checkout remains outside this prototype.

Cosmetic guidance, not a medical diagnosis. Product prices are a reference snapshot and may change. VTO colors are approximate inspirations, not certified brand shade matches.

Built With

• css
• javascript
• node.js
• react
• vercel
• vite
• youcam-makeup-vto-api
• youcam-skin-analysis-api