All credits to eldritchadam
This lightweight LoRA brings a consistent feel of richly textured genuine oil painting. With a blended combination of various Western traditions, this LoRA style feels familiar and timeless.
Extremely versatile, you can use ClassipeintXL with just the base model without the refiner, write straightforward prompts and consistently get the image you ask for. I'm sharing straight generations with no upscaling or retouching, and almost always only prompted once or twice with 2 generations per prompt, so there was little cherry-picking. The output is just extremely consistent.
Update: after more testing I have identified one substantial artifact - a tendency of the model to exaggerate head sizes and body proportions, leaning toward child-like or dwarf-like dimensions in some situations. I should be able to correct for this in v1.1
Most of the images I share here were at a full 1.0 model strength. I only reduce to about 0.8 if I want to get a recognizable portrait of a famous person. Full strength, the LoRA will make people less recognizable, but outside of that one scenario, I prefer the full strength for the richness and texture it brings.
And most of my prompts are extremely plain, simple language describing the scene.
As of the day I'm sharing it, the model is still somewhat new to me, so I've put it through paces enough to know I'm wholly satisfied with its versatility, ease of use, and that it captures the aesthetic I was after. But I have yet to push it to strange places and I look forward to seeing what more it can do. I know it has surprises in store! I hope you'll share some images and discoveries with me.