WearLync - AI Virtual Try-On Platform
AI-powered SaaS platform enabling fashion brands to generate photorealistic try-on imagery using advanced generative AI models.
Project overview
An AI-powered SaaS platform for fashion imagery generation at scale.
WearLync empowers fashion brands and retailers to generate photorealistic try-on imagery at scale. By combining AI models (FASHN VTON v1.5, Google Vertex AI) with an intuitive platform, teams can replace expensive physical photoshoots with on-demand, diverse imagery generation. The platform includes credit-based billing, custom model training, admin dashboards, and async processing pipelines.
What it does
WearLync enables fashion brands to upload garment images, select AI models and scenes, then generate photorealistic try-on visuals using FASHN VTON v1.5 and Google Vertex AI—eliminating expensive physical photoshoots.
Problem solved
Replaces time-consuming, costly physical photoshoots with on-demand AI generation. Enables diverse, inclusive model imagery at scale without logistics overhead.
Main goal
Democratize high-quality fashion imagery generation, allowing brands of all sizes to create compelling visual content rapidly and affordably.
Tech stack
Practical tools chosen for reliability and speed.
Frontend
- Next.js 16 + React 19
- TypeScript
- Tailwind CSS v4
- Framer Motion
- shadcn/ui
Backend API
- Node.js/Express 5
- TypeScript
- Prisma ORM
- JWT Auth
- Zod Validation
AI/ML
- FASHN VTON v1.5
- Google Vertex AI
- Lambda Workers
- GPU Processing
Integrations
- Razorpay Payments
- Cloudinary Storage
- AWS Lambda
- Redis Queue (BullMQ)
Infrastructure
- AWS (EC2→ECS/Fargate)
- PostgreSQL
- Aiven Redis
- Docker & NVIDIA Toolkit
Security
- Plan-tier Gating
- Credit Ledger System
- Audit Logging
- Role-based Access Control
Build process
A focused delivery path from problem mapping to launch.
Discovery
Analyzed fashion industry workflow, photoshoot costs, model diversity requirements, and brand imagery needs.
Planning
Defined features: AI generation, credit billing, custom models, catalog management, admin dashboard, async queue.
Core Build
Built Next.js frontend, Express API, Prisma schema, JWT auth, credit system, and Razorpay integration.
AI Pipeline
Integrated FASHN VTON v1.5 and Google Vertex AI; implemented Lambda worker for async processing.
SaaS Layer
Added subscription tiers (FREE, PAID_TRIAL, PRO, BUSINESS), trial flows, and admin credit management.
Custom Models
Built training pipeline allowing users to create brand-specific AI models from reference images.
Deployment
Prepared AWS hosting, CloudWatch monitoring, SSE status streaming, and ECS/Fargate migration strategy.
Challenges and solutions
The important problems solved for the business.
Problem
Worker Architecture
Solution
Migrated from EC2-hosted Node worker to Lambda for scale-to-zero and cost efficiency while maintaining parity.
Impact
Achieved reliable async processing with CloudWatch monitoring and fallback Python GPU worker.
Problem
Design System Drift
Solution
Consolidated three separate visual systems (marketing, dashboard, prototype) into unified design tokens and components.
Impact
Reduced maintenance burden and improved user trust through consistent, professional interface.
Problem
Auth Model Evolution
Solution
Migrated from legacy combined role enum to split planTier + accessRole with backward compatibility.
Impact
Enabled fine-grained permissions (SUPPORT, ANALYST, SUPER_ADMIN) while maintaining existing routes.
Problem
Real-time Status Updates
Solution
Implemented SSE endpoint for live generation job status alongside polling fallback for reliability.
Impact
Customers receive instant feedback without overloading the API with polling requests.
Problem
Credit Ledger & Billing
Solution
Built immutable transaction log supporting 7+ transaction types (purchase, generation cost, refunds, grants, expiry).
Impact
Accurate financial tracking, audit trails, and transparent customer billing visibility.
Results and outcome
A complete SaaS foundation with measurable operational gains.
50%
faster content creation vs. physical photoshoots
70%
cost reduction in imagery production
100+
compatible fashion brands
15 sec
average generation time per image
8+
subscription tiers and billing scenarios
7
AI model providers integrated
Next case study