Is it for Krea 2 or Z-Image? I tested both it didn't do anything. I guess I got trolled.

I like it. Its great for getting things more bulky.

For those following my LoRA hints and tricks: This is a style LoRA – basically the endboss of LoRA training. The model has to guess which elements in the training images actually define the style. Pop‑culture styles with different people, outfits and backgrounds make it hard for the model to latch onto a single consistent feature.

I ran strength tests and noticed that hip‑hop related details only start to appear around 1.5 strength. Mostly jewelry becomes larger and more frequent. Usually LoRA‑related changes begin around 0.18 and are clearly visible around 0.60, so the fact that this LoRA only shows its style markers at 1.5 suggests it has very specific detail genes. The common element in the training set was probably jewelry.

To counter this “detailer effect”, I added a detail‑enhancing LoRA at –1 strength. The idea: If I subtract general style‑detail amplification, maybe the hip‑hop‑specific parts get pushed forward. Surprisingly, this worked quite well for me.