VRGDG_LUTS

VRGDG_LUTS is a one-stop color-grading node: feed it an image (or a batch of video frames - an IMAGE tensor is an IMAGE tensor), pick a look from a dropdown of 11 built-in .cube LUTs, and out comes the graded result. This is the "LUT loader" the pack's README had on its roadmap, now shipped.

For an AI music-video workflow this is quietly important. AI video comes out of the model flat and neutral; it's the grade that makes it feel like a music video instead of a tech demo. Being able to drop the same LUT across every scene chunk is what gives a whole multi-scene video one coherent look instead of forty slightly different color moods.

The inputs that matter

Most of the node is one dropdown plus a strength slider.

• lut_name - pick from 11 bundled .cube files: cinematic ones (FGCineDrama, FGCineWarm, FGCineBright), filmic ones (Filmic_Bright_65, Filmic_Dark_65, Cool_Bright_65, Vintage_Fade_65), and a set of "Presetpro" emulations (Fuji Provia, Cine Grade, Ektachrome, Moody Grade). These ship with the pack, so there's nothing to download.
• strength - 0 to 10, default 10 (full). Lower it to blend the graded image back toward the original; 5-ish is a good start for "grade, but don't relight the scene."
• device - auto, cuda, or cpu. Leave it on auto; it picks the GPU when available. You only touch this if you're debugging a crash.

The image output is the same shape as the input, so this node slots anywhere between a decoder and a video save node - put it after the VAE decode and before the preview/save.

The mechanism, briefly

A .cube LUT is a lookup table that remaps colors - the same format video editors and DaVinci Resolve have used for years. Applying it here is a per-pixel remap on the tensor, and because it's one pass over pixels, it's cheap enough to run on long videos without melting your GPU. That's the whole trick: no model, no inference, just a color transform with an opacity blend on top.

Getting it

Same pack install as everything else:

cd ComfyUI/custom_nodes
git clone https://github.com/vrgamegirl19/comfyui-vrgamedevgirl

or ComfyUI Manager → search "vrgamedev", then restart. kornia from the pack's requirements is the dependency that does the heavy lifting here.

Where people get tripped up: strength maxes at 10, not 1. If you're coming from nodes where strength means 0.0–1.0, a "10" that looks fully applied isn't broken - that's full strength by design. And if the LUT looks too strong at default, slide strength down before blaming the LUT choice; the same .cube at strength 4 vs 10 is a completely different look. For a beginner this is probably the most accessible node in the pack - instant, no models, and a genuinely useful way to make a generated video feel shot rather than rendered.