Stun
AI-powered shade discovery for iOS ~ using TrueDepth AR face scanning and social media extraction to eliminate makeup guesswork across 10,000+ shades.
Services
iOS App + AI Development
Platforms
iOS
Date
2025
Challenge
The beauty industry sells over 10,000 shades across 1,000+ brands, yet most people still pick foundation by holding a bottle next to their wrist under store lighting. The result is billions in returns, wasted product, and consumers who settle for "close enough" because there's no reliable way to match shade to skin.
On top of that, millions of beauty routines go viral on TikTok and Instagram daily, but the products featured are matched to the creator's skin — not the viewer's. There was no bridge between seeing a shade you like on screen and knowing whether it would actually work for you. We needed to build something that could analyze skin with clinical precision, cross-reference a massive product database in real time, and plug directly into the social platforms where people actually discover beauty products.
Solution
We built Stun as a native iOS app using Apple's TrueDepth camera and ARKit to perform a 30-second face scan. The system maps the user's face with a 3D wireframe, tracks facial geometry in real time, and runs AI analysis on undertone, depth, and surface characteristics. That scan produces a skin profile that gets matched against a database of 10,000+ shades from over 1,000 brands, each result scored with a confidence percentage so users know exactly how close the match is.
The second piece was social integration. Users paste a TikTok or Instagram link and Stun's AI extracts every product from the beauty routine, then cross-references each shade against the user's skin profile. Instead of guessing whether a creator's foundation works on a different skin tone, Stun returns personalized alternatives ranked by match accuracy. All matched shades feed into a cloud-synced vanity collection that users can organize, share with friends, and access across devices. Internal testing showed 99.8% accuracy on shade matching across select datasets.
Let's talk about your project!
