TryOn AI - try on clothes

Instantly try on any outfit, dress, or style on your own photo with photorealistic AI. See how clothes look on you before you buy.

Comment

Inspiration

We’ve all been there: you see an outfit online—on Instagram, Pinterest, or an e-commerce store—and wonder, "Will this actually look good on me, or does it just look good on the 6-foot studio model?"

Online clothing returns are notoriously high (often 30–40%), largely because people can't visualize how fabrics, cuts, and colors will fit their actual body and skin tone. We wanted to build a genuine, zero-friction virtual fitting room that puts anyone in any outfit in seconds—just from a quick phone photo and a clothing screenshot.

What it does

TryOn AI lets you upload a photo of yourself and a picture of any outfit—from a leather jacket or summer dress to traditional wear like sarees and suits.

In under 10 seconds, the app generates a realistic photo of you wearing the outfit:

  • Preserves Identity: Your face, hairstyle, skin tone, and body proportions remain 100% natural.
  • Realistic Fabric Physics: It captures folds, textures, zippers, embroidery, and lighting rather than doing a flat sticker overlay.
  • Adaptive Framing: If you upload a selfie but the outfit is a full-length gown, the AI intelligently pulls back the virtual camera to show the complete outfit.

How we built it

  • Mobile App: Built with React Native (Expo) with native performance modules for both iOS and Android.
  • Backend & Cloud: Google Cloud Functions (2nd Gen) and Firebase Firestore & Storage for serverless scaling.
  • AI Core: Powered by Google's Gemini multimodal model (gemini-3.1-flash-image-preview) with structured prompt chains.
  • Live Tuning: A custom Firestore dynamic config layer and web admin dashboard that lets us tune AI framing and prompts on live traffic without requiring app updates.

Challenges we ran into

  1. Camera Framing & Close-up Selfies: Users often upload close-up selfies, but choose full-length outfits. Early on, the AI would crop the garment. We solved this by developing an adaptive outpainting prompt system that automatically expands the frame and generates natural body anatomy when full garments are detected.
  2. Complex Garments & Layering: Outfits like sarees, lehengas, or multi-piece suits required strict segmentation rules so the AI completely replaces existing clothes rather than awkwardly blending old jeans under a traditional dress.
  3. Speed & Latency: Balancing maximum image resolution with sub-10s generation times on mobile connections required careful client-side compression and serverless memory tuning.

Accomplishments that we're proud of

  • Achieving photorealistic try-on results that look like real camera photos instead of synthetic AI art.
  • Handling diverse clothing categories (casual tops, jackets, 2-piece suits, and complex traditional Indian attire).
  • A sub-10 second turnaround time that feels smooth and engaging on a real mobile device.

What we learned

  • Generative AI prompt engineering for visual tasks is deeply iterative—treating prompts like dynamic configuration rather than hardcoded strings was essential to rapid improvement.
  • Edge cases in human photos (angles, lighting, partial poses) require clear fallbacks and adaptive framing logic.

What's next for TryOn AI

  • Mix & Match Styling: Letting users try on separate tops, bottoms, and accessories together in a single custom outfit.
  • Direct Store Integration: Chrome extension and shopping integrations to try on clothes directly on brand websites before hitting "Buy Now".
  • Video Try-On: Short motion clips to see garments in movement.

Built With

Share this project:

Updates

Submission history