AI Beauty Color Lab

Inspiration

Most beauty advice online is not truly personal. People rely on trends, influencers, or endless trial-and-error, yet still struggle to find what actually complements their face, skin tone, and natural coloring. We wanted to build something more precise — a tool where beauty recommendations are driven by data, not guesswork. The vision was to combine AI-powered color analysis with real-time AR try-on, so users could move from "what should I wear?" to "here is exactly how it looks on me" in seconds.

What it does

AI Beauty Color Lab transforms a single selfie into a fully personalized beauty experience. The user uploads a photo, and the system analyzes their skin tone, undertone, and facial features to determine their personal color season (warm, cool, deep, or light). From that analysis, it generates a tailored "signature look" — a curated combination of shades and styles matched to the user's natural coloring. The user can then try the look on their own face instantly through augmented reality, adjusting and exploring variations in real time. Every recommendation is data-backed and visually testable, removing the guesswork from finding what truly suits each person.

How we built it

We built the platform around two core engines: Gemini AI for color analysis and aesthetic reasoning, and YouCam AR for realistic makeup rendering. Gemini interprets the user's facial features and translates them into structured attributes — tone, undertone, contrast level, and seasonal classification — which feed into a recommendation layer that maps these traits to specific product shades and styles. YouCam AR then renders the recommended look on the user's face in real time, ensuring the AI's suggestions are not just theoretically correct but visually convincing. We modeled personalization as a function:

Look = f(skin tone, undertone, facial features, style preference)

This formulation gave us a consistent framework to generate, test, and refine recommendations across diverse users.

Challenges we ran into

The biggest technical challenge was aligning AI-generated recommendations with what actually looks accurate in AR. Small differences in lighting, skin tone interpretation, or color mapping could meaningfully change how realistic a rendered look appeared, so we invested significant time in normalizing outputs and testing across a wide range of faces and lighting conditions. The deeper challenge, though, was conceptual: translating subjective beauty language into structured data. Concepts like "natural glow" or "soft glam" had to be decomposed into measurable attributes such as tone, saturation, and intensity. Bridging that gap between human aesthetic intuition and machine-readable parameters required constant iteration between design judgment and engineering rigor.

Accomplishments that we're proud of

We are proud of building an end-to-end pipeline that takes a user from selfie to fully rendered, personalized look in real time. Getting Gemini's analysis to feed cleanly into YouCam's AR engine — and producing results that genuinely look right on diverse skin tones — was a significant integration achievement. We are also proud that the system feels less like a novelty filter and more like a real personalization tool: the recommendations are explainable, the AR results are believable, and the experience holds up across different users and conditions.

What we learned

The hardest part of AI in beauty is not generation — it is translation. Turning human aesthetic judgment into structured, consistent outputs is where the real value is created. We learned that aligning two complex systems (an AI reasoning engine and a real-time AR renderer) requires carefully designed intermediate representations, not just good models on either end. We also learned how much rigor is required to make subjective domains feel objective: every "vibe" has to be backed by parameters, and every parameter has to be tested against real human perception.

What's next for AI Beauty Color Lab

This project is a foundation, not an endpoint. Next, we plan to expand the system into a full AI-driven beauty discovery and shopping experience — connecting personalized looks directly to real products, building a feedback loop where users can refine recommendations over time, and broadening the model to cover hair color, accessories, and full styling. Longer term, we see Color Lab evolving into a personalized beauty operating system: one where every recommendation is data-backed, every product is visually testable in real time, and every user gets advice tailored to who they actually are rather than to whatever is trending.

Built With

• node.js