This is a model trained in a style similar to Clash Royale, mainly focused on character design.
It's currently performing well in terms of art style fidelity, but still shows structural issues in some areas—such as hands and weapons.
For better control and accuracy, using tools like ControlNet or image-to-image (img2img) is recommended when possible.
Usage Tips:
Prompting:
For best results, try using short phrases and modular keywords when writing prompts.
A suggested format is: trigger word + subject + style + background + quality.
This helps the model generate more accurate and stylistically consistent outputs.Base Model:
sd_xl_base_1.0 URL：https://huggingface.co/stabilityai/stable-diffusion-xl-base-1.0/tree/main
And of course, feel free to try different XL models—they can generate some truly amazing results!Different models may respond to prompts in unique ways, so it’s worth experimenting to see what works best for you.
LoRA Weight:
0.7 ~ 1.0
CFG Scale:
7 ~ 10
Resolution:
1024x1024
Samplers:
Recommended: DPM++ 2M SDE
Scheduler: Karras
Steps: 28-30
Hires. Fix Settings:
Upscaler: R-ESRGAN 4x+
Denoising strength: 0.3 ~ 0.45
If you like this model, you can buy me a coffee https://ko-fi.com/lightscape. I would truly appreciate your support and it will motivate me to keep creating.