◎ Radiance Apply Grade Info

The most annoying thing about getting a color grade right is that it's stuck to one image. Every shot you generate needs the same look re-applied, by hand, hoping you remember the numbers. RadianceApplyGradeInfo exists to kill that workflow: it takes a saved grade - a JSON string produced by Radiance's grade system - and replays it onto any image, with a strength slider to dial the look in or out.

The mechanics are clean. The grade_info string carries the full parameter set (per-channel lift/gamma/gain, offset, contrast, pivot, saturation), and this node parses it and runs the same _apply_grade math the main grade node runs. So "replicates a RadianceGrade exactly" isn't marketing - it's literally the same code path, per the source. Feed it a grade you saved from one generation and it produces a frame that matches, shot to shot.

The inputs

• image - the new shot you want graded.
• grade_info - a STRING, force-input (it's a socket, not a text box). Connect this to the grade_info output of a RadianceGrade node, or paste a saved JSON grade. The tooltip is unambiguous: "Connect to the grade_info output of RadianceGrade."
• strength (0–1, default 1) - blends between the original (0) and the fully graded (1) image. This is your "how much of the look" dial, and it's the reason this node is better than just re-grading by eye.

Outputs: image (the graded frame, float32) and grade_info (the same JSON passed through, so you can chain or branch the look).

The workflow it unlocks

This is the batch-consistency node. Generate a reference shot, spend time getting the grade right in RadianceGrade, save the grade_info, then run every subsequent generation through Apply Grade Info with that string. Same look, no re-grading, and a strength slider for "same grade, but subtler on this one." If you've ever wanted a colorist's "grade the hero, match the rest" pipeline inside ComfyUI, this is it. It pairs with the README's v2.3.3 mention of Grade Match for statistics transfer - this node is the replay side of that coin.

Caveats

• The grade is a set of lift/gamma/gain/offset values in a working space, so it assumes the input images are in the same color space and exposure range the grade was designed for. Grade a batch of similar shots and it's magic; grade an SDR PNG with an HDR-targeted look and it'll fight you.
• Invalid JSON in grade_info is handled gracefully (the source logs and returns the image untouched), but a valid but wrong grade just grades wrong - garbage in, wrong look out.

Install

Part of Radiance - ComfyUI Manager → search Radiance, or:

cd ComfyUI/custom_nodes
git clone https://github.com/fxtdstudios/radiance.git
cd radiance
pip install -r requirements.txt

Restart. Under FXTD Studios/Radiance/Color.