The latest articles on DEV Community by varshini ramesh (@varshini_ramesh_649d54dd5). https://dev.to/varshini_ramesh_649d54dd5 https://media2.dev.to/dynamic/image/width=90,height=90,fit=cover,gravity=auto,format=auto/https:%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Fuser%2Fprofile_image%2F4171020%2Fbcfa5202-a0c8-461e-bdb8-883b5943953a.png https://dev.to/varshini_ramesh_649d54dd5 en varshini ramesh Thu, 08 Oct 2026 10:55:21 +0000 https://dev.to/varshini_ramesh_649d54dd5/glamsync-3cgl https://dev.to/varshini_ramesh_649d54dd5/glamsync-3cgl <h2> GlamSync </h2> <p>What if your laptop camera could do more than just show you your reflection?</p> <p>We built <strong>GlamSync AI</strong>, a personal beauty styling coach that combines <strong>Gemma 4, OpenCV, and real-time AR</strong> to help users understand what makeup and styling choices actually suit them.</p> <h2> The Problem </h2> <p>Choosing makeup that works with your individual face, skin undertone, and outfit can be surprisingly difficult.</p> <p>Most beauty applications either provide static filters or generic recommendations. They may tell you <em>what</em> to use, but they don't understand your facial features or show you <em>how</em> to apply it.</p> <p>We wanted to bridge that gap.</p> <h2> Our Solution </h2> <p>GlamSync AI creates a personalized styling pipeline:</p> <p><strong>Outfit → Facial Analysis → AI Recommendation → AR Application Guidance</strong></p> <h2> 1. Outfit Color Analysis </h2> <p>We use OpenCV and K-Means clustering to extract dominant colors from an outfit and understand its palette.</p> <h2> 2. Facial Analysis </h2> <p>OpenCV YuNet analyzes facial geometry while CIELAB-based ITA colorimetry and skin segmentation help estimate skin tone and undertone.</p> <h2> 3. Gemma 4 Reasoning </h2> <p>The user's facial profile, outfit colors, occasion, budget, and available products are passed to <strong>Gemma 4</strong>, which generates personalized styling looks and step-by-step recommendations.</p> <h2> 4. Real-Time AR Coach </h2> <p>This is where GlamSync goes beyond a normal recommendation system.</p> <p>Using <strong>MediaPipe's 468-point Face Mesh</strong>, our AR mirror tracks facial landmarks and overlays guidance for areas such as the cheeks, eyelids, eyeliner, and lips.</p> <p>The user can follow the guidance while applying their makeup, with voice instructions for a hands-free experience.</p> <h2> Why Open-Source AI Matters </h2> <p>We wanted AI to be an actual part of the product rather than simply adding a chatbot.</p> <p>Our pipeline combines open/open-weight AI and local computer vision components so that facial analysis and personalization are tightly integrated into the application.</p> <p>This also gives us the flexibility to experiment with different models and reduce dependence on a single closed AI provider.</p> <h2> Tech Stack </h2> <ul> <li>React + TypeScript + Vite</li> <li>FastAPI + Python</li> <li>Gemma 4 Multimodal</li> <li>OpenCV + YuNet</li> <li>MediaPipe Face Mesh</li> <li>NumPy</li> <li>CIELAB / ITA color analysis</li> <li>Google AI Studio</li> <li>WebRTC / Web Speech API</li> </ul> <h2> What We Built </h2> <p>The final system takes the user from:</p> <blockquote> <p><strong>"What suits me?"</strong></p> </blockquote> <p>to</p> <blockquote> <p><strong>"What should I use?"</strong></p> </blockquote> <p>to</p> <blockquote> <p><strong>"Show me exactly how to apply it."</strong></p> </blockquote> <p>That's what we wanted GlamSync AI to be — not just an AI beauty recommender, but an interactive <strong>AI styling coach</strong>.</p> <h2> Built for Hacktoberfest Hack Day Coimbatore </h2> <p>Built by Team <strong>GlamSync AI</strong> for Hacktoberfest Hack Day Coimbatore.</p> <p>🔗 <strong>GitHub:</strong> <a href="https://github.com/guruc1234a-spec/hacktoberfest-hack-day-coimbatore-x-init-club-and-idea-club" rel="noopener noreferrer">https://github.com/guruc1234a-spec/hacktoberfest-hack-day-coimbatore-x-init-club-and-idea-club</a></p> <h2> 🎥 Demo Video </h2> <p><a href="https://youtu.be/WVJn0bj-v3U" rel="noopener noreferrer">▶️ Watch GlamSync AI in action on YouTube</a></p> ai opensource python api