LUT Generation

If you've ever graded an image, applied the same grade to fifty more, and then had to do it all over again on the next batch, you know the pain this node is aimed at. LF_LUTGeneration takes a color transformation dataset - the per-color before/after mapping produced by the Color Analysis node - and bakes it into a lookup-table dataset you can reuse, so the same grade stops being a one-off tweak and becomes a repeatable recipe.

Let's be clear about what it is and isn't. It's an Analytics-category node, meaning its job is to derive and expose data, not to render a final image. The single required input is color_analysis_dataset, typed as JSON - the transformation dataset generated by the Color Analysis Node (LF_ColorAnalysis) from the same pack. The Color Analysis node looks at an image, works out how its colors should shift to hit some target look, and emits that shift as a dataset. This node converts that dataset into lut_dataset, a JSON lookup table: source colors mapped to their transformed versions. That output is what you'd feed into a color-grading node downstream to apply the same transformation to other images consistently.

The whole thing is deliberately small - one input, one output - because it's a middle link in a pipeline, not a standalone tool. You'll notice it's flagged as an output node, so its natural home is the end of a data chain where the resulting LUT dataset is your deliverable: save it, hand it off, or just inspect it.

Two honest caveats. First, this only makes sense inside the LF Analytics family - if you don't use the Color Analysis node, this has nothing to chew on; there's no fallback input for a raw image. Second, "LUT" here is a dataset in LF's JSON format, not a .cube file you can drop into Photoshop. It plays within this pack's ecosystem rather than the broader color-grading toolchain. That's a real limitation if you were hoping to export to an external editor.

Install is the standard LF Nodes path:

cd ComfyUI/custom_nodes
git clone https://github.com/lucafoscili/comfyui-lf

or via ComfyUI Manager by searching "LF Nodes", then restarting. No extra models or dependencies - this is pure data processing on CPU. And the pack-level note applies: comfyui-lf is in legacy mode (frozen, still working), with development now in lucafoscili/lf-nodes, where the same node family lives.

For a first run: load an image, run it through Color Analysis with a target look, connect the color_analysis_dataset output here, and read the lut_dataset. If you're iterating on a look, that's your repeatable starting point.