Color: ColorLab (ACR)

This is the node that makes the pack's "pro-grade color suite" claim believable. JHPixelProColorLab is a monolithic Adobe Camera Raw-style color node: one node body with 55 parameters that run a deterministic Basic → HSL → Color Grading → Gray Mix pipeline, all in pure PyTorch, all on the tensor. If you've ever wished ComfyUI had a proper grading tool instead of a pile of math nodes, this is the one that says "yes, but it's one tall node."

The pipeline order matters and it's worth knowing: Basic first (global exposure and tone), then HSL (per-color-family adjustments), then Color Grading (shadow/midtone/highlight tinting), then Gray Mix (black & white weights). Everything defaults to neutral, and all-zero with gray_enable=False is a true identity - so it's safe to drop into a workflow and dial up from zero, knowing you're starting from "nothing changed."

What's behind the 55 knobs

The parameter names group into panels:

• Basic (basic_*, 11 controls): exposure, contrast, highlights, shadows, whites, blacks, texture, clarity, dehaze, vibrance, saturation. This is your exposure/tone pass in one cluster.
• HSL (hsl_*, 24 controls): eight hue anchors - red, orange, yellow, green, aqua, blue, purple, magenta - each with hue, saturation, and luminance. This is the selective-color panel; if you've used the pack's JHPixelProHueSaturationRange and wanted it for all families at once, this is the all-at-once version.
• Color Grading (grade_*, 12 controls): shadow, midtone, and highlight hue / saturation / luminance plus balance per region - the teal-and-orange cinematic look lives here.
• Gray Mix (gray_*, 8 controls): gray_enable plus seven B&W mix weights. Skipped entirely when disabled, so leaving it off costs you nothing.

There's a basic_* input for every one of these plus the image input, and a single image output. That's the whole interface.

The honest caveats

Two things to internalize before you get frustrated. First, there is no custom tabbed widget - all 55 controls are scrollable in one node body, grouped by naming convention. It works, but it's a wall of sliders; treat the basic_ / hsl_ / grade_ / gray_ prefixes as your navigation. Second, this is an ACR-compatible approximation, not Adobe's code - the tone masks, HSL hue falloff, and grading masks are pure PyTorch approximations. For a retouch/grading pass it's excellent; for pixel-perfect reproduction of a specific Lightroom preset, it won't be identical.

There's a practical play here worth stealing from the pack's own workflow library: run your grade through ColorLab, then use the HALD pipeline (JHPixelProHALDIdentity → chain → JHPixelProLUTExport) to bake that wall of sliders into a portable .cube LUT. Grade interactively in ColorLab, ship the result as a LUT.

Install

ComfyUI Manager → search ComfyUI-JH-PixelPro, or:

cd ComfyUI/custom_nodes
git clone https://github.com/jetthuangai/ComfyUI-JH-PixelPro.git
cd ComfyUI-JH-PixelPro
pip install -r requirements.txt

Restart ComfyUI; it's under ComfyUI-JH-PixelPro/color. The pack needs ComfyUI ≥ 0.43.x, Python ≥ 3.10, plus kornia, mediapipe, opencv-python-headless, and scipy.

My verdict: this is the node people will either love or dismiss on sight. The love case is real - it's the only thing in this pack that gives you a full ACR-style grade in one place, which is exactly what "professional retouching in ComfyUI" was missing. The dismissal case is also fair: 55 scrollable sliders is a lot of interface for a graph tool. Start with Basic and one HSL family, verify with a before/after, and let the node earn its height.