This repository provides custom nodes for ComfyUI, enabling real-time webcam and screen capture for generative AI workflows. It targets users of Stable Diffusion, particularly those leveraging fast models like LCM/SDXL Turbo, by offering seamless integration of live visual input for iterative image generation and manipulation.

How It Works

The core functionality revolves around custom nodes that interface with system-level capture mechanisms. It includes nodes for direct webcam capture, screen region capture, and window capture, feeding these live inputs into ComfyUI. Additional nodes facilitate image processing, such as depth-to-normal map conversion, noise removal, and visual area masking, enhancing the utility of the captured data for advanced prompting and control.

Quick Start & Requirements

• Install by cloning the repository into the ComfyUI/custom_nodes directory:

  git clone https://github.com/toyxyz/ComfyUI_toyxyz_test_nodes
  

• Run setup.bat within the cloned directory for webcam capture dependencies.
• Requires ComfyUI and Python. Specific hardware requirements for real-time processing depend on the ComfyUI workflow.
• Documentation and example workflows are available within the repository.

Highlighted Details

• Real-time webcam and screen capture nodes for ComfyUI.
• Nodes for visual area masking, random text loading, GLB export, noise removal, and depth-to-normal conversion.
• Includes utilities for saving images, exporting videos, and face detection.
• Offers direct screen capture with options for full screen, window, or relative window capture.

Maintenance & Community

The project appears to be actively developed with recent updates noted in the README. Community interaction channels are not explicitly mentioned.

Licensing & Compatibility

The repository does not explicitly state a license. Users should verify compatibility for commercial use or closed-source integration.

Limitations & Caveats

Window capture stability can vary depending on the target application. Some features, like AI Render overlay, are noted as unstable and require caution. The project recommends using alternative tools for ControlNet preprocessing to avoid performance bottlenecks.