Back
Year
2023
Tech & Technique
React.js, Python, Hugging Face, Civit AI LoRA, Tailwind CSS, Vercel
Description
A generative art platform built for WhoJoshi using Stable Diffusion models to create stylized AI artwork based on user prompts. Integrated with Hugging Face-hosted models and LoRA fine-tunings for high customization.
Key Features:
Technical Highlights:
Key Features:
- 🎨 Prompt-to-Image Generator: Users can generate AI art using custom text prompts
- 🔧 LoRA Switching: Dynamic application of different LoRA fine-tuned models for varied styles
- 🧠 Hugging Face Integration: Hosted and called inference endpoints securely
- 📱 Responsive Design: Optimized interface for mobile and desktop users
- ⚡ Fast Generation Feedback: Asynchronous task queue to handle image generation efficiently
Technical Highlights:
- Integrated Hugging Face Inference API with token-based authorization
- Built Python backend to manage prompt processing, LoRA selection, and image output
- Designed frontend with React and Tailwind for a smooth UX
- Implemented asynchronous queue system (using Celery + Redis or similar) to manage inference load
- Added local image gallery with download and share options
My Role
Full-Stack Developer
Owned end-to-end development and model integration:
Owned end-to-end development and model integration:
- ✅ Backend: Created a Python-based server to handle Stable Diffusion prompt processing and LoRA selection
- 🎨 Frontend: Built the UI in React with Tailwind CSS, optimized for art previews
- 🧠 Model Ops: Integrated multiple LoRA models via Hugging Face endpoints
- 🔁 Async Workflow: Set up task queue system for smooth async generation (Celery/Redis)
- 🚀 Deployment: Deployed backend on AWS EC2 and frontend on Vercel
- 📂 File Handling: Managed image output storage and delivery via secure download links