A fine-tuned Style LoRA for the Z-Image (Base) model, trained on a full curated dataset of Vincent van Gogh masterworks at native 1024x1024 resolution. Designed to bring rich impasto oil textures, dynamic swirling brushstrokes, and iconic color palettes to both classic subjects and modern/cyberpunk concept art.
Recommended Settings
Base Model
Z-Image (Standard / Non-Turbo)
Model Weight / Strength
0.40 (optimal range: 0.35 – 0.45)
CLIP Weight
0.40
Architecture / Rank
Rank 128 / Alpha 128
Dataset Size
800 high-resolution curated artworks (2,000 steps)
Prompting Advice for Z-Image
Z-Image requires descriptive, structured prompts to unlock the full potential of this LoRA. For style transfer onto sci-fi, vehicles, or architecture, explicitly specify background materials and brushstroke styles in your text prompt.
Example Prompt Template:
van gogh style, thick impasto oil painting, heavy palette knife brushstrokes, vibrant colors, [YOUR SUBJECT HERE], detailed mechanical armor, glowing lights, swirling starry sky background, masterwork, 8k
Description
Comments (4)
What are the .json files for? Are they ComfyUI-specific? Also, your model card is still showing Flux2 at the moment.
No, not necessarily it’s just part of the workflow. Thanks, I hadn't noticed that flux2 was in the description 😅.
That’s what sleepless nights do to you 😅😉
I did exactly the same with a recent LoRA - just trying to get it posted quickly so I could go to bed 😂