Verso

See it, style it, sustain it Your AI mirror for sustainable fashion. Try on looks virtually before buying to reduce returns, waste, and environmental impact

Comment 1

presentation link: https://docs.google.com/presentation/d/1JG8oVNZL0kApwltY5OzDlrelVaDg-4-mVtfcYjVRmCg/edit?usp=sharing

About the Project

Inspiration

The fashion industry is one of the largest contributors to environmental waste, with millions of tons of clothing ending up in landfills each year. But the problem isn't just overproduction - it's uninformed purchasing decisions. Research shows that online clothing purchases have return rates ranging from 20-30% , creating a massive carbon footprint from shipping and processing. For apparel specifically, the average return rate is 24.4%, nearly 8 percentage points higher than the overall online return rate . Beyond returns, impulse purchases lead to closets full of barely-worn items.

We saw a massive market opportunity at the intersection of sustainability and consumer confidence. By helping people make smarter, more confident decisions before they buy, we can reduce waste while building a scalable business model. Our app lets users virtually try on different styles, outfits, and accessories using AI, empowering them to explore new looks without the environmental and financial cost of returns, fast fashion cycles, and impulse purchases.

This isn't just good for the planet - it's a multi-billion dollar problem waiting to be solved.

What It Does

Our app allows users to:

  • Upload their own photo
  • Swap facial attributes, hairstyles, hair colors, and eye colors using AI
  • Recolor clothing items to visualize different outfit combinations
  • Make more informed, sustainable fashion choices

Think of it as your personal AI stylist and fitting room - try before you buy, reducing waste and boosting confidence.

Market Opportunity

The global virtual try-on market was estimated at $9.17 billion in 2023 and is projected to reach $46.42 billion by 2030, growing at a CAGR of 26.4% . This growth is driven by rising return rates and consumer demand for personalized shopping experiences. Our solution addresses multiple pain points:

  • For consumers: Reduce purchase anxiety, minimize returns, discover new styles risk-free
  • For retailers: Return processing costs are substantial, with the apparel industry facing an estimated $25 billion in processing costs
  • For the planet: Measurable reduction in fashion waste and carbon emissions

Our revenue model includes B2C subscriptions for premium features, B2B partnerships with sustainable fashion brands, and affiliate commissions from successful purchases made after virtual try-ons.

  • ** for the backend server and API orchestration
  • ** for hair, eye, and facial feature transformations with masking flows in the backend
  • Google Gemini for outfit top/bottom region detection using polygon tracking
  • Local sharp pipeline (non-AI) for final fast top/bottom recolor execution
  • React for the frontend interface and user experience
  • CSS for responsive, modern styling
  • Base64 image processing to handle user uploads and AI transformations

The core workflow:

  1. User uploads their photo
  2. Selects the style attributes they want to try (hairstyle, hair color, eye color, clothing colors)
  3. Our AI pipeline processes facial features through Replicate
  4. Gemini detects clothing regions via polygon tracking
  5. Local pipeline executes fast recoloring of detected clothing regions
  6. User can experiment with multiple looks before deciding what to actually purchase

What We Learned

  • Multi-model AI pipelines can combine the strengths of different services - using specialized models for specific tasks produces better results than relying on a single solution, while keeping infrastructure costs manageable
  • The market validation is strong: Users care about sustainability, but they care even more about making confident decisions - framing our app around "smart shopping" creates a compelling value proposition that drives user adoption
  • Polygon tracking for garment detection is more reliable than simple segmentation for varied clothing styles, giving us a technical moat
  • B2B potential is massive: Early conversations with fashion retailers revealed that return rates are their biggest profit drain - our technology could save them millions
  • Balancing image quality with processing speed is crucial for good UX and unit economics - we optimized by using fast local pipelines for final color transforms
  • Handling privacy and image data responsibly is critical for building user trust and long-term brand value

Challenges We Faced

  • Getting AI to work accurately on specific attributes: Our initial vision was to allow users to upload any style picture and extract specific attributes from that image to apply to their own photo. However, due to time constraints and limited resources, we had to pivot to a predefined attribute selection approach. This pivot actually revealed a more scalable product strategy - curated attribute libraries that we can expand systematically.
  • Clothing color accuracy: While we successfully implemented a good range of hair colors, hairstyles, and eye colors, the color swapping for clothes is not perfect. The local recolor pipeline works quickly but doesn't always handle complex patterns, shadows, or fabric textures as smoothly as we'd like. We see this as our biggest technical challenge to solve for achieving product-market fit.
  • Region detection complexity: Getting Gemini's polygon tracking to consistently identify top and bottom garment boundaries across different poses, body types, and clothing styles required significant tuning. This taught us valuable lessons about model selection and data preprocessing.
  • Model integration: Coordinating between Replicate's masking flows, Gemini's detection, and our local pipeline meant managing multiple API calls and ensuring proper image format conversions at each step. Building a robust, scalable backend architecture was crucial.

Business Model & Go-To-Market Strategy

Phase 1 (Current): Free consumer app with basic features to build user base and gather data

Phase 2 (6 months): Premium tier ($9.99/month) with unlimited try-ons and advanced features

Phase 3 (12 months): B2B partnerships with sustainable fashion brands - white-label our technology for their e-commerce sites, charging per API call or licensing fee

Phase 4 (18 months): Affiliate revenue from purchases influenced by our platform

Our competitive advantages:

  • Multi-model AI pipeline that balances quality and cost
  • Focus on sustainability creates authentic brand differentiation
  • Privacy-first approach builds consumer trust
  • Scalable B2B model with clear ROI for retail partners

What's Next

Short-term (3 months):

  • Improve clothing color accuracy by refining our recolor algorithms to better handle patterns and textures
  • Launch beta program with 1,000 users to gather feedback and usage data
  • Begin pilot partnerships with 2-3 sustainable fashion brands

Medium-term (6-12 months):

  • Expand to the original vision: allow users to upload inspiration photos and extract specific style attributes
  • Add more facial feature options and accessory overlays
  • Implement a "save your looks" feature and social sharing for viral growth
  • Scale to 50,000+ users and prove conversion metrics

Long-term (12-24 months):

  • Raise seed funding to expand team and accelerate development
  • Build out full B2B platform for enterprise retailers
  • Partner with major sustainable fashion brands for direct shopping integration
  • Expand internationally, starting with European markets where sustainability regulations are driving demand

We're not just building an app - we're building the future of sustainable fashion commerce.

Sources

  • National Retail Federation & Happy Returns (2024). Customer Returns in the Retail Industry Report
  • Coresight Research (2024). Online Apparel Return Rates Study
  • Grand View Research (2024). Virtual Try-On Market Report

Built With

Share this project:

Updates