Painterly 3D-to-2D Style — SDXL LoRA v03
A style LoRA trained toward painterly color blocks, hand-painted textures, three-dimensional forms, and cinematic lighting. This is v03, trained for 1,200 steps.
Suggested settings
Base: SDXL Base 1.0
Trigger: pntv03 style
LoRA strength: 0.6–0.8; start at 0.7
Sampler: DPM++ 2M Karras
Steps: 30; CFG: 5.5
Resolution: start around 1024-pixel resolution
Gallery images
The gallery images were supplied by the publisher and generated using Z-image. They are visual style references and do not represent tested outputs of this SDXL LoRA.
Training
SDXL U-Net LoRA; rank/alpha 32; 1,200 steps; 30 source images plus 10 detail crops; resolution bucketing around 1024.
Usage
Results depend on the base model, prompt, seed, and LoRA strength. Inspect faces, hands, and complex props. Higher strengths may affect composition.
License
The adapter weights are released under CC BY-NC 4.0 for noncommercial use. The SDXL base model’s CreativeML Open RAIL++-M restrictions also apply.
Description
v03, 1,200 training steps. SDXL Base 1.0 U-Net LoRA; rank/alpha 32. Trigger: pntv03 style. Start at strength 0.7 (suggested 0.6–0.8), 30 steps, DPM++ 2M Karras, CFG 5.5. Gallery images are publisher-supplied Z-image style references, not tested SDXL LoRA outputs.