Adobe TechMeet Submission – AI Image Editing & Generation
This repository contains our solution for the Adobe TechMeet challenge: a dual-component system delivering professional-grade AI image editing and generation tools on the web.
NebulaEdit is designed with a decoupled architecture: a fast, interactive frontend and a robust backend for heavy AI inference. The system hides the complexity of ComfyUI nodes behind a familiar, layer-based editing interface.
Our research highlights that most existing tools trade off control for simplicity. NebulaEdit targets creators who need precise, localized edits (e.g., relighting, element fusion) rather than generic image generation.
For a detailed look at the system architecture and execution pipeline, see:
galaxy-canvas-ai/: React + Vite frontend. Provides the canvas, toolbars, and interaction logic for features like MagicQuill and Image Fusion.nebula-server/: Python/FastAPI backend. Bridges to ComfyUI, managing complex node workflows for AI features (Qwen-Image-Edit, Relighting, etc.).setup.sh: Deployment script for the server. Installs ComfyUI, Python venv, and fetches all required model weights from Hugging Face.
Below are key screenshots demonstrating NebulaEdit's features and user interface:
These screenshots illustrate the workflow from uploading an image, choosing an edit type, entering prompts, and viewing the results of advanced AI-powered editing features.
The following images from the wireframes demonstrate the design and planned user experience of NebulaEdit:
These wireframe screenshots provide insight into the application's design philosophy and user-centric interface planning.
See setup_instructions.md for detailed setup and deployment steps.















