This is a rebuild of my other Natural Noob XL VPRED model, now based on NoobAI EPS and Illustrious models. A large handful of popular and obscure checkpoints and LoRas have been repeatedly merged into the mix using my own semi-random merging system and then I hand-picked the best result at each step using a series of tests to ensure a fairly consistent quality.

The goal is a checkpoint that knows a lot of subjects, characters and styles, while generally having an aesthetically pleasing look.

Usage Tips:

For the best balance between color and prompt accuracy I like a CFG between 4.0 to 6.0 and 24-36 steps, but the model can handle a CFG range of about 3.0 to 7.0 without destabilizing too much.

Example images larger than the standard SDXL resolutions use hi-res fix so the actual total step count for those images is double what is listed.

I usually begin my positive prompts with something like:

very awa, masterpiece, absurdres,

Followed by style and artist names such as:

traditional media, oil painting \(medium\), absurdres,

And then whatever characters, poses, objects, viewpoints etc.

Use a mixture of danbooru & e621 tags, and very simple sentences.