ComfyUI custom nodes for image enhancement via detectors, detailers, upscalers
Top 18.5% on sourcepulse
This repository provides a comprehensive suite of custom nodes for ComfyUI, aimed at enhancing image generation workflows with advanced detection, detailing, upscaling, and masking capabilities. It targets users seeking granular control over image manipulation and refinement within the Stable Diffusion ecosystem.
How It Works
The pack integrates various segmentation and detection models (SAM, CLIPSeg, ONNX, MediaPipe) to identify and isolate image regions. These identified regions (SEGS) can then be processed by specialized "Detailer" nodes for targeted refinement, upscaling, or other modifications. The architecture supports complex workflows through pipe nodes, conditional logic, and advanced sampling techniques like iterative upscaling and regional sampling.
Quick Start & Requirements
ComfyUI/custom_nodes
and running pip install -r requirements.txt
within the cloned directory.segment-anything
, scikit-image
, piexif
, opencv-python
, scipy
, numpy<2
, dill
, matplotlib
(optional). Specific Linux packages (libgl1-mesa-glx
, libglib2.0-0
) may be required.ComfyUI-Impact-Subpack
is required for Ultralytics Detector Provider nodes.Highlighted Details
Maintenance & Community
Licensing & Compatibility
Limitations & Caveats
UltralyticsDetectorProvider
requires a separate subpack installation.Control Bridge
node has limitations with batch count > 1 and specific execution order requirements.2 days ago
1 day