I wrote out the prompt with variables in it, so I could change the variables or get them automatically filled in by randomly picking from lists. See https://github.com/adieyal/sd-dynamic-prompts#combinatorial-generation and the part about variables in the extension’s help.

But the real thing that gives the images structure is the regional prompter extension, which interprets the ADDROW/ADDCOL stuff and makes the one thing be on one side of the image and the other thing be on the other side, by doing different parts of the image with different prompts in the same generation.

OK, I tried it via the maintained fork here, and it does something, but it doesn’t really let you zero out a LoRa. Here is the model not understanding my prompt at all and drawing garbage, and then here it is when I increase the LoRa weight in the prompt from zero and it draws something different. In both cases I am using the extension to tell the LoRa not to come in until step 21 of my 20 step run. In both cases I told the extension to plot the LoRa weight it thinks it is using, and it was 0 at all steps. But clearly having the LoRa in there did something even when it was supposed to be at weight 0.

What do you mean it’s phallic? 😂

Hm. I’ve never really figured out AND, but this might be useful. Thanks!

img2space

This was several different ones; it should be embedded in each image, but I think cyberrealistic_v31 worked well.

I tried it with LoRas; it works OK but the LoRa applies over the whole image, and I found that the LoRas seem to make the model less good at reading comprehension, which is what I am fighting with the regional prompter in the first place.

Really what I want is to use the LoRa for some generation steps but not use it for others, but I haven’t worked out how to do that yet.

Bonus:

These models make such amazing non-clothes.

I mean, it kind of is? Not everybody has the same number of legs, so the model should turn out people with just the one as often as the real world does, unless the prompt tells it different.

Never seen anybody with quite so many knees though.

These folks are great, I want them in Skyrim now.

How are they doing their “buy model” thing? They aren’t selling checkpoints, they are taking one image and like dreamboothing the person they generated so they can then make more images that are recognizably them.

This is a really great idea.

These are done with the A1111 web UI.

This is done on a 4080, which is pretty quick and can do a batch size of 6 or maybe 8 of these at a time. You can definitely run the model on worse cards, or even on CPU, at slower speeds, smaller image sizes, or fewer images per batch.

And you don’t actually need any hardware; you can use frontends for AI Horde and borrow other people’s hardware instead.

You do need to be comfortable with a page full of sliders and boxes that you have no idea what they do, though. I don’t understand most of the settings either.