BustySophieXL版本v1.1 (ID: 1032500)

Use this Lora to place Sophie into the scene of your choice. Built against ZavyChromaXL, this model works best with that checkpoint but will work with any XL model including illustration models like Cheyenne.

My favorite generation settings are:

• euler a /polyexponential

• 30-60 steps (usually 30-40 is fine)

• CFG 5-8 (usually 6)

• any dimensions as long as one side is at least 768 pixels

The "euler a" sampler seems to like higher steps and CFG in order to produce good results, especially with background details. The more you bake it, the better it gets.

The sample pictures were all generated with very simple clothing prompts. No resizing, no adetailer, no extra Loras. I've played with some XL Loras and they appear to mix pretty well with Sophie, I just didn't want to use them as samples of the base character.

The girl has been a recurring character in images generated with my VoluptuousMix merge checkpoint (https://civitai.com/models/618381/voluptuousmix-v10) and I wanted to be able to use her in XL images, so I created this Lora with some very careful upscaling. I think she turned out pretty well.

Edited to add: This model is intended for SDXL checkpoints, but it also works with XL Pony models. You may need to boost the strength above one (1.2 to 1.5) to ensure the face and hair are correct. Don't forget to boost both the Lora itself and the keyword "bustysophie" in your prompt for best results.

Enjoy!

描述:

I discovered the "Resize LoRA" tab in Kohya_ss. The bustysophie models have always been unusually large, but this tool has made them MUCH more manageable through the magic of mathematics. I could have gone even smaller, but not without losing some of the facial accuracy. This new version produces results that are close to identical to v1.0 in most of the images I've tested.

Smaller size means less room on your hard drive, but also much faster load times when you're creating images. Enjoy!

训练词语: bustysophie