A cinematic color-grade / reality-enhancer look. A STYLE LoRA for MiniMax H3 video - it changes the look, not the subject.
EXPERIMENTAL: first-run conversion, lightly tested - the sample clips on this page are the only test so far. Strength 1.0 is a good starting point for this photoreal look. Epoch 35 is the softer file, epoch 50 the stronger. Tell me what you find.
Trigger word: cinemagradestyle (put it first in the prompt). Strength 1.0. Two files: epoch 35 and epoch 50 - try 35 first, 50 for a smoother finish.
Base: MiniMax H3 (fl2va). Loads in ComfyUI with LoraLoaderModelOnly (the file must sit in models/loras/).
How it was made: 150-image Krea 2 dataset generated with the source style LoRA, every subject unique, 50% people, cinematic shot vocabulary in every caption; trained with Fizgig "MiniMax H3 Style (LoRA 8)" - rank 8, LR 2e-4, 50 epochs, blocks 0-3,6-47, adamw, shift 0.6667.
Best results: give H3 a pencil-sketch storyboard plus character references and let the LoRA fill in the look. With a photoreal start frame it mostly reinforces what is already there.
Samples: a pencil-sketch storyboard written for THIS style, rendered by H3 with two characters from this LoRA's own dataset as the only references - so what you see is the LoRA filling in the world. H3 renders at 480p, 6 s, seed 77, audio stripped.
Credit: derived from "CinemaGrade - Krea 2 Reality Enhancer" by nottheninja (Krea 2) - https://civarchive.com/models/2902077/cinemagrade-krea-2-reality-enhancer?modelVersionId=3281617 - please support the original creator. Shared with nottheninja's permission, free and non-commercial.
Description
FAQ
Comments (3)
is there a way to apply this to an existing video clip eg reference video?
Am I doing something wrong, or it falls apart with slightly faster movement than sitting? Characters limbs and clothing parts fly around at 1 str or 0.6 str equally. I got a semi-decent result at 0.5 but how much of the LoRA is preserved at that str.
His models are very bad.
Multiple models published that all looks the same, you can see when someone is just doing this for clicks and doesn't care what the actual results look like.