II

A more flexible model. Supports many more tags and concepts than my previous model, and is trained at 768-based bins for better large, detailed images.

Dial-a- capability anywhere from maximum realism on up past ridiculous. Can technically make images with no at all, but who would want to?

This model isn't intended to be e.g. a good or model. Instead, it's designed to play nice with other models and whatever big prompts, by just adding this model and a few select keywords.

Also, I tried to make a very broad model with this one but there are plenty of focused models that do a more narrow concept better or more reliably.

Instructions

Short answer: just run it at like 0.4-0.8 and add "" to the prompt somewhere.

Long answer: Click the 🛈 info button at the lower right of any example image (or just click the image) to see details of how it was made. Most of my example images have the prompt etc. included, so flip through the examples and copy what you like. I've added a list of tags that specifically exist in the dataset, but you can also just add any -related language to your prompt and your generations should become spunkier. Using a list of tags is more robust than just one, and adding weight to the whole list is an effective way to scale the volume up or down. You can turn the tags up and the main lora weight down to get similar results without interfering as much with concepts in other loras or the base checkpoint.

Tags/Activation

Training images were auto-tagged with Booru tags and a few concepts that the captioner didn't reliably pick up on were manually tagged as well. These tags are more likely to be understood by both the lora and the base model. Here's an alphabetical list of the -related tags that appear in at least one dataset image:

• after

• after

• covered in

• in

• in clothes

• in mouth

• in

• in shoes

• on

• on body

• on boy

• on

• on

• on clothes

• on feet

• on floor

• on hair

• on hair

• on hand

• on hands

• on stomach

• on

• on

• pool

• string

• swapping

• cumdrip

• excessive

• mouth full of

• projectile

• suggestive fluid