Inspiration
Celebrity endorsements drive a 59% increase in purchase intent and a 4% boost in conversions — but they cost six figures and take weeks to produce. Meanwhile, independent brands and small e-commerce sellers are stuck with flat product photos on white backgrounds, unable to compete with the aspirational imagery that moves product.
We asked: what if any brand could hire a supermodel for a photoshoot in under 5 minutes, for under $150? And what if the models and celebrities whose likenesses power these AI generations actually got paid for it?
Lookbook was born from the intersection of two ideas — democratizing high-end product photography through AI, and building an ethical framework where real models consent to and earn royalties from every generation that uses their likeness.
What it does
Lookbook is a platform where anyone can create professional, photorealistic product campaigns using AI supermodels — and the models get paid for it.
Here's how it works:
- Pick a package — Choose from three tiers (LookBook at $149, Digital Presence at $249, or Professional Catalogue at $899), each with different model counts, product limits, and image deliverables.
- Browse the Model Marketplace — Select from verified AI models with detailed profiles, ratings, and portfolio galleries. Each model has a pricing tier (basic, premium, elite) and earns a royalty on every image generated with their likeness.
- Upload your products — Add up to 5 images per product. We support bags, clothing, hats, scarves, and shoes — each routed to the correct YouCam AI try-on API automatically based on your category selection.
- Watch the magic happen — Our AI generates photorealistic try-on images in real time. You see completed images appear as a live thumbnail strip while the rest are still processing.
- Export everywhere — Download all images, regenerate any you want to tweak, or export directly to your Shopify store as fully formed products with images, descriptions, and inventory — ready to sell.
How we built it
Frontend: React 19 + TypeScript with Vite, styled with Tailwind CSS and animated with Framer Motion. State management uses Zustand with localStorage persistence so campaigns survive page refreshes. The model marketplace features real-time filtering, profile modals, and a multi-image product uploader.
Backend: Python FastAPI handling image processing, task orchestration, and third-party integrations. Pillow handles automatic format conversion (AVIF, WebP → JPEG) since the AI APIs only accept JPEG/PNG.
YouCam/PerfectCorp APIs: This is the core engine. We integrated multiple try-on endpoints from the YouCam Enterprise API:
- Bag Try-On — person image + bag, with gender and style parameters
- Clothing Try-On — supports full body, upper body, and lower body garment categories
- Scarf Try-On — person image + scarf, with gender-aware generation
- Hat and Shoes Try-On — same accessory pattern with category-specific routing
AgenticAI : Kilocode CLI was a gamechanger for us.
The flow for each: upload images to PerfectCorp's File API via presigned URLs → create a try-on task → poll every 2 seconds until completion → cache the result locally (since S3 presigned URLs expire after ~2 hours).
Shopify Integration: OAuth 2.0 client credentials flow with automatic token caching and refresh. When you hit "Export to Shopify," we create real Shopify products with the original product photo + all AI-generated images, HTML descriptions, pricing, and inventory — ready to appear on your storefront.
Challenges we ran into
Image format hell. PerfectCorp's APIs only accept JPEG and PNG, but modern browsers and cameras love AVIF and WebP. We had to build an auto-conversion pipeline that detects the actual image format (not just the file extension) and converts on the fly.
Reversed parameters. Not all YouCam APIs work the same way. The clothing try-on API expects src=clothing, ref=person (the opposite of every other endpoint where src=person). Debugging why every clothing generation returned garbage took longer than we'd like to admit.
S3 URL expiry. PerfectCorp returns results as presigned S3 URLs that expire after about 2 hours. If a user came back to download their images later, they'd get 403s. We solved this by caching every result image locally on the backend the moment it completes.
Category-specific routing. Each YouCam API has slightly different parameter requirements — bags need gender + style, clothing needs garment_category, scarves and hats need gender. Building a unified generation pipeline that routes to the right endpoint with the right parameters based on product category was a real architectural puzzle.
Accomplishments that we're proud of
- Multi-category try-on fully integrated with smart routing — bags, clothing, scarves, hats, and shoes each call the correct YouCam API with the correct parameters automatically.
- Real-time generation UI — users see their images appear one by one as a live thumbnail strip while polling continues in the background. It feels like watching a photoshoot happen.
- End-to-end Shopify pipeline — from uploading a product photo to having a fully published Shopify listing with AI-generated lifestyle imagery, in minutes.
- Ethical AI model marketplace — every model in Lookbook is verified, consents to AI usage, and earns fair royalties. We built the pricing model so that model compensation is baked into every tier.
- Single-image regeneration — don't like one shot? Click regenerate, and we'll create a fresh variation using the same model and product, without re-running the entire campaign.
What we learned
Working with the YouCam/PerfectCorp API suite taught us that AI try-on isn't one-size-fits-all — every product category has different requirements for source/reference image orientation, parameter sets, and result quality. Building a clean abstraction over multiple endpoints with varying contracts was a crash course in API design.
We also learned that the gap between "AI generates an image" and "this is a production-ready e-commerce workflow" is massive. Handling format conversion, result caching, presigned URL expiry, Shopify OAuth, and real-time polling — the plumbing around the AI is where most of the engineering effort goes.
Finally, we learned that ethical AI and good business aren't at odds. Building model compensation into the pricing model from day one actually makes the product more compelling — brands feel better about the imagery they're using, and models gain a new passive revenue stream.
What's next for Lookbook
- Expanded model marketplace — onboarding real models and celebrities who can license their likeness, set their own rates, and track earnings through a creator dashboard.
- AI styling assistant — use AI to analyze products and recommend model pairings, poses, and campaign themes automatically.
- Multi-platform export — beyond Shopify, support WooCommerce, Amazon, and social media formats (Instagram carousels, TikTok product cards).
- Campaign analytics — track which AI-generated images drive the most engagement and conversions, and feed that data back into smarter generation.
- Video generation — extend from still images to short-form product videos using the same ethical AI framework.
Built With
- fastapi
- python
- react
- tailwind
- typescript
Log in or sign up for Devpost to join the conversation.