Boobkake22 ee9d798ec7 Remove deprecated Node.js 20 publishing actions
Use actions/checkout@v7 and actions/setup-python@v7, which run on Node.js 24.
Invoke the official Comfy CLI directly because the upstream publishing
composite still embeds older setup actions.

Preserve the existing publish triggers, Python version, registry secret,
package version and read-only repository permissions. Pass the registry
token through a quoted environment variable.

Validation: actionlint and shell syntax checks pass; publish triggers and
secret references match the previous workflow. No registry publication is
triggered by this workflow-only change.
2026-09-09 22:27:16 -07:00
2026-05-26 04:21:27 -07:00
2026-05-26 04:21:27 -07:00
2026-06-10 04:35:36 -07:00
2026-05-26 04:21:27 -07:00

ComfyUI Color Correct GPU

A small ComfyUI custom node that mirrors the controls and processing order from Art Venture's ColorCorrect node while doing the image math with vectorized PyTorch ops.

Node

Color Correct GPU

Category: ArtVenture/Post Processing

Inputs match the Art Venture node:

  • image
  • temperature
  • hue
  • brightness
  • contrast
  • saturation
  • gamma

The node ID is ColorCorrectGPU, so it can be installed beside comfyui-art-venture without colliding with its existing ColorCorrect node.

Notes

  • This avoids the upstream CPU path that converts each frame through NumPy, PIL, and OpenCV.
  • The GPU path intentionally keeps the upstream 8-bit quantization steps and PIL/OpenCV-style color math so results stay close to the original node.
  • Neutral saturation and hue settings skip their color-space conversions to avoid unnecessary work and tiny no-op conversion drift.
  • If the incoming ComfyUI IMAGE tensor is on CPU, the node moves it to ComfyUI's torch device when one is available, then returns it to ComfyUI's intermediate device for compatibility.
  • Results should visually match the original node, but may not be byte-identical because PIL/OpenCV and torch can differ slightly in floating-point edge cases.
S
Description
No description provided
Readme
31 KiB
Languages
Python 100%