OpenViz is an AI-powered design application that transforms hand-sketched drawings and imported images into photorealistic renders using advanced text-to-image models. Optimized for designers, architects, and product creators, OpenViz provides a fast, intuitive workflow to explore design iterations visually in seconds.
- Konva-powered Engine: Smooth, high-performance canvas interaction.
- Drawing Tools: Brush, Eraser, and Shape tools (Rectangles, Circles, Lines) with configurable properties.
- Symmetry Mode: Mirror your drawings vertically or horizontally for perfectly balanced designs.
- Viewport Controls: Smooth zoom (10% - 500%) and panning.
- Full Control: Reorder, rename, hide, or delete layers.
- Opacity & Blending: Adjust layer transparency and blend modes (Normal, Multiply, Screen, Overlay).
- Live Thumbnails: Real-time visual feedback for every layer in the sidebar.
- Type-specific Layers: Support for sketch, image, and render layers.
- ComfyUI Integration: Connect directly to your local ComfyUI instance for generation.
- Prompt Engineering: Multi-line prompt input with real-time character counting.
- Style Presets: Quickly switch between "Photorealistic", "Cyberpunk", "Watercolor", and more.
- Influence Slider: Control exactly how much the AI should follow your sketch vs. your prompt.
- Batch Generation: Generate multiple variations in one go.
- Modern Dark UI: A sleek, glassmorphic interface designed for productivity.
- Custom Color Picker: Detailed color selection with HEX, HSL, and SV support.
- Responsive Layout: Optimized for desktop with a fluid, adaptive panel system.
- Node.js (v18+)
OpenViz uses ComfyUI as its background rendering engine. To ensure a smooth connection:
-
Launch ComfyUI with the following flags to allow the Vite proxy to communicate with it:
python main.py --listen --port 7821
(Note: If you use the default port 8188, update
targetUrlinvite.config.tsaccordingly). -
Required Models: The default workflow expects the following models in your
ComfyUI/modelsdirectory:- Checkpoints:
sd3.5_large_fp8_scaled.safetensors(incheckpoints/) - ControlNet:
sd3.5_large_controlnet_canny.safetensors(incontrolnet/)
- Checkpoints:
-
WSL Users: If you are running OpenViz in WSL and ComfyUI on Windows:
- The
vite.config.tsis pre-configured to automatically find your Windows host IP. - Ensure your Windows Firewall allows inbound connections on the ComfyUI port (7821).
- The
-
Clone the repository:
git clone https://github.com/Sander-HR/openviz.git cd openviz -
Install dependencies:
npm install
-
Start the development server:
npm run dev
- Basic canvas interaction and drawing tools.
- Layer system with thumbnails and opacity.
- ComfyUI integration for image generation.
- Advanced color picker and theme refinement.
- Refine Mode: Iterative refinement of generated renders.
- Import/Export: Support for importing external images and exporting projects as PNG/JPG or
.openvizJSON. - Shape Transformations: Move, scale, and rotate objects after placement.
- Workflow Presets: Save and load custom ComfyUI workflows within the UI.
- Workbench switch working on a larger canvas with multiple images and render modes.
- AI Prompt Generator: Semi-automated prompt crafting based on sketch contents.
- Mobile Optimization: Touch-optimized UI for tablets and styluses.
- Cloud Sync: Cross-device project synchronization.
- Framework: React 18
- State Management: Zustand
- Canvas Engine: Konva.js
- Styling: Tailwind CSS + Lucide Icons
- Build Tool: Vite
- AI Interface: ComfyUI API (WebSocket/HTTP)
This project is licensed under the MIT License - see the LICENSE file for details.