AI Avatar
Stable Diffusion & Dreambooth web app with model training, image generation, GFPGAN upscaling, and gallery management.
Overview
StableFusion is a full-stack web application for AI image generation using Stable Diffusion and Dreambooth. Users can upload selfies to train custom models, generate images with text prompts, and enhance outputs with GFPGAN upscaling. The React/Redux frontend follows atomic design (Environment > Pages > Organisms > Molecules > Atoms) with Bulma CSS. The Express/MongoDB backend handles authentication via Passport.js with email verification, image upload/download, model training orchestration, and inference API routes. Deployed with Docker Compose.
Challenge
Users needed an accessible web interface for training custom Stable Diffusion models and generating AI images without command-line tools or GPU setup.
Results
50,000+ avatars generated in the first month, 88% user satisfaction, and sub-30s average generation time.
Key Highlights
Architecture
Timeline
Backend & Auth
3 weeksExpress server, MongoDB/Mongoose, Passport.js auth, email verification routes
AI Pipeline
6 weeksStable Diffusion inference, Dreambooth training, GFPGAN upscaling, prompt API
Frontend
5 weeksReact/Redux atomic design, Bulma UI, image upload/gallery, studio & training pages
Deploy & Polish
2 weeksDocker Compose, image download/export, testing with Jest
Related Projects
Animatrix AI
AI-powered image and video generation SaaS with Dreambooth training, element extraction, Prisma ORM, and Stripe subscription billing.
Arty AI
AI art e-commerce platform with NestJS backend, social auth (Apple, Facebook, Google, Twitter), i18n, and print-on-demand checkout.
Karandia
Real-time video platform with WebRTC meetings, live broadcasts, screen sharing, recording, and a video editor — powered by a NestJS microservices backend.
Ready to Transform Your
Business with AI?
Get a free consultation and discover how AI can automate your operations, reduce costs, and drive growth.