⭐StableMondAI-SDG⭐

A merge to bring back SDXL art styles to noobaiXL which resulted in a stabilised model that is perfect for toony style generations.

Prompt recommendations
Generation recommendations
Sampler/Scheduler comparison
Style examples
Artist examples
Example Generations

🟨Model Description

Developed by: Big Lasagna
Funded by: /sdg/
Merged from: lasagnavpredreal, noobAIXL, SDXL
Direct Download link
Also available on civitai.com
Also available on tensor.art
Artist tag gallery: rentry.org

🟨Uses

This model is an attempt to make the recent noobAI XL based merges better understand the art styles that were present in the SDXL 1.0 base model but got lost in the lineage of style merges. This has worked somewhat but the main benefit is that it stabilised noobaiXL towards better coherence and a more stable default style, while still being very flexible and responsive to style and artist tags.

Photorealism is not a strong point of this model, check out the other MondAI models if you are looking for that.

This model is in V-PREDICTION mode.

Prompts need to use danbooru/e621 tags, no Natural Language Support is added at this point. Use the common quality tags at your own discretion, for some style mixes they improve and for others they deteriorate the generation quality.

See below for examples and comparison grids.

🟨Prompt Recommendations

Start with a basic prompt and work from there. All IllustriousXL/noobAIXL quality tags are recognized but these models work better with leaner prompts.

Starter Positive prompt:

masterpiece, best quality, newest,

Starter Negative prompt:

worst quality,

🟨Generation Recommendations

The following settings are recommended starting points:

Sampler: Euler A
Scheduler: SGM Uniform
Steps: 26-30
CFG: 4-5

RescaleCFG is not required but 0.7 works to make the colors pop slightly less on some mixes.

Many sampler and scheduler combinations work well, the SDE samplers need specific schedulers and the Karras scheduler is not recommended.

🟨More Information

🟨Quantized version

A kind anon created quantized versions of this model which you can use if you are very low on VRAM or for special setups like Mixmod. You can find them in the quantized model subfolder of the download section.