PerfectSkinDiary

Track your skin. Prove your routine works. PerfectSkinDiary uses AI to turn daily selfies into professional skin scores with a personalized goal image to keep you motivated.

Comment

PerfectSkinDiary

Inspiration

As someone who was born and raised in Taiwan, we take great care of our skin. In December last year, I used various products incorrectly, which damaged my skin barrier and resulted in Post-Inflammatory Erythema (PIE).

When I first encountered this challenge, I knew I had to participate because it resonated deeply with my situation. I wanted to create an app that could track and help users monitor their skin's progress. While users can observe their skin changes by looking in the mirror daily, it's challenging to track these changes consistently. AI can assist in this by utilizing the expertise of Perfect Corp's professional lab team and equipment to provide numerical data that is easier to interpret and understand your skin health.

Moreover, AI can help you assess whether your current morning and evening skincare routine is effective. It generates PDF reports that you can share with your dermatologist, providing accurate data from Perfect Corp's lab team.

What it does

PerfectSkinDiary is a mobile app for AI-powered skin tracking. Users create a Track for a specific goal — a stubborn acne cluster, a product trial, a general monthly baseline — and log one photo per day.

On Day 1, the app does two things in parallel:

  • Establishes a baseline using Perfect Corp's YouCam AI Skin Analysis (HD), measuring 16 skin dimensions including moisture, redness, texture, pores, wrinkles, radiance, and more
  • Generates a Goal Image using AI Skin Simulation — a photorealistic render of the user's face with their targeted concerns improved, giving every tracking journey a concrete, motivating destination.

Each subsequent daily entry is automatically compared against the previous one, surfacing delta scores (↑ moisture +8, ↓ redness -3). Claude translates the raw scores into a 2–4 sentence plain-language summary, contextualised by the AM/PM skincare routine the user has logged for that Track.

When ready for a dermatologist consultation, users can export 3, 5, or 10 days’ selection of entries as a structured PDF — photos, mask overlays, scores, and summaries — sorted chronologically.

A UV Index widget on the home screen recommends the minimum SPF for the day based on the user's location. (SPF feature in the app is something that people forget often. Everyone should wear sunscreen every day to keep their skin healthy.)

How I built it

Frontend: React Native with Expo and TypeScript, using Expo Router for navigation. Platform-specific UI files (.ios.tsx / .android.tsx) give iOS a native Liquid Glass design (iOS 26, expo-blur BlurView) while Android uses a custom warm diary palette defined in our design system.

Backend: Supabase provides the PostgreSQL database, file storage, anonymous auth, and Edge Functions (Deno) as a secure API proxy. All sensitive API calls — to Perfect Corp and Anthropic — are routed through Edge Functions.

Perfect Corp APIs used:

  • AI-Skin-Analysis (HD) — 16-metric daily skin scoring with regional breakdowns and mask overlays
  • AI-Skin-Simulation — Day 1 goal image generation (selected concerns at intensity 0.5)

LLM: Claude with prompt caching for cost-efficient structured JSON-to-summary tasks. Edge Function supports both single-entry summaries and multi-entry trend insights.

Accomplishments that I proud of

  • The Goal Image concept. Giving users a photorealistic, locked north-star image on Day 1 — generated from their own face. It transforms abstract score tracking into a personal, visual journey.

  • Two APIs, two distinct moments. Each Perfect Corp API is used at a different lifecycle stage: Skin Simulation on Day 1, and Skin Analysis every day thereafter.

  • Professional PDF export. The PDF export produces a chronologically ordered clinical record that a dermatologist can actually use, not just a vanity report.

What I learned

  • Start with dummy data. I used up all the 1000 units from Perfect Corp because HD analysis requires 20 units per run. Should start with a dummy data and reuse that for developing.

What's next for PerfectSkinDiary

  • Monetization

  • Privacy: Photos are encrypted client-side before upload to Supabase Storage. Even the developer cannot view user photos.

  • Notifications Notify user to apply sunscreen and record photo for skin analysis if they haven't done so for a while.

  • Product recommendation Generates a list of recommendations of skin products to help users solve their target concerns.

Built With

Share this project:

Updates