---

title: Aurora AI emoji: 🌌 colorFrom: purple colorTo: pink sdk: docker app_port: 7860 pinned: false

Aurora AI

Aurora AI is a modern, web-based AI image editing studio for background removal, image enhancement, and background replacement. The project is built with a clean separation between a Next.js frontend and a FastAPI-based ML inference backend.

Overview

Aurora AI provides an intuitive, Canva-inspired interface for applying AI-powered image edits directly in the browser. All machine learning inference is handled by a dedicated backend, allowing the frontend to remain fast, responsive, and UI-focused.

The project is structured for clarity, extensibility, and production-style development, with frontend and backend concerns clearly separated.

Features

• Image Enhancement
  • AI-powered upscaling
  • 2x (Fast) and 4x (High Quality) modes
• Background Removal
  • One-click background removal
  • Transparent PNG outputs
  • Enables downstream background replacement
• Background Replacement
  • Replace backgrounds using uploaded images
  • Automatically gated until a transparent image is available
• AI Background Generation (Experimental)
  • Prompt-based background generation
  • Integrated into the background replacement workflow

Architecture

Aurora AI follows a clean, production-oriented frontend/backend split.

Frontend

• Next.js (App Router)
• React
• TypeScript
• Tailwind CSS
• Responsive, tool-driven UI with progressive disclosure

Backend

• FastAPI for REST APIs
• PyTorch for ML inference
• Modular inference pipeline for image processing tasks

Models & Inference

• BiRefNet for background removal
• ESRGAN / DAT-based models for image upscaling
• Diffusion-based models for AI background generation
• Automatic device detection (CPU / GPU where available)

Project Structure

.
├── backend/          # FastAPI backend
│   ├── app/         # API entry point and routing
│   ├── models/      # ML model integrations
│   └── data/        # Input/output image data
├── frontend/        # Next.js frontend application
└── README.md

Getting Started

Backend

cd backend
pip install -r requirements.txt
uvicorn backend.app.main:app --reload

The backend API will be available at http://localhost:8000.

Frontend

cd frontend
npm install
npm run dev

The frontend will be available at http://localhost:3000.

Docker Deployment

Single Container Deployment

Aurora AI can be deployed as a single Docker container that runs both the frontend and backend services.

Configuration:

• SDK: docker
• App Port: 7860

Build and Run:

docker build -t aurora-ai .
docker run -p 7860:7860 aurora-ai

The application will be available at http://localhost:7860.

Environment Variables:

• PORT=7860 - Port for the Next.js frontend (default: 7860)
• HF_HOME=/tmp/.huggingface - Hugging Face cache directory (optional, for free tier)

Example with custom port:

docker run -p 8080:8080 -e PORT=8080 aurora-ai