This Is Amazing.

Who’s a good puppy? 🐶💕

He is a very precious bean. :3

Blessing your feed.

How I make AI pics

Several people have asked me how I'm creating my sissy AI pics, either what tools or what prompts I'm giving the AI. I thought I'd lay out the process that works for me.

Tools

I'm using the Easy Diffusion distribution of Stable Diffusion. It's an AI image generator designed to run locally on a PC, rather than on a central cloud server. Because of this it gets around some of the limitations on the types of images you can create (NSFW and ABDL are particularly tricky). The downside of this is that you need a PC with a fairly meaty graphics card to run the software.

As well as the basic Stable Diffusion install, I've also downloaded several "models" which are sets of data that train SD to work with certain images. There are models for realistic-looking people, manga and so on. I've had a lot of success with the Deliberate, HardBlend and ABDL models available from civitai.com.

Process

Basically the process is a lot of trial and error. I usually start off with a prompt something like this:

Full body portrait photo of (((Justin Bieber))) wearing short frilly crop-top with big puff sleeves and panties. Glamourous pink makeup. Bonnet. Pigtails. Detailed and intricate. (((((Sad))))) unhappy (((((shy))))) (((embarrassed))) scared. Eyes closed. Long eyelashes. Mouth slightly open. Standing In a blue kitchen.

I let the system throw out some images and then refine the prompt. I often use names of famous people because the AI models have been trained on those faces - Justin Bieber seems to work well, although I have no particular like or dislike of Bieber! The final image doesn't always end up looking like that person.

Also important is the negative prompt, the things you DONT want to see in the final image. My negative prompt has built up in time to look like this:

Deformed, blurry, bad anatomy, disfigured, poorly drawn face, mutation, mutated, extra limb, ugly, poorly drawn hands, missing limb, blurry, floating limbs, disconnected limbs, malformed hands, blur, out of focus, long neck, long body, ((((mutated hands and fingers)))), (((out of frame))) (((baby))) (((toddler))) (((child))) (((doll))) tattoo (((happy)))

After quite a bit of tinkering, I'll get four images something like this:

I'll then often take a pic that looks promising and then get the AI to generate four more using that picture as a basis:

This was using a model called cyberrealistic, which is good for some types of clothing and photo-realism, but not trained with diaper and ABDL pics. There are other models that are trained with ABDL pics but aren't very good at photo-realism or variations of clothing. The trick here I use is to redo a section using a different model, using "inpainting". First selecting the area to be in-painted...

Then choosing another diaper-trained model to in-paint the missing section, based on what was there already...

That gives a variety of different underwear options for the poor sissy.

It's looking good, but I want to sissify-up that top by about 50%. To do that I literally draw on top of the image with a finger-painting of what I want...

It doesn't matter that the painting looks really crude, the AI will just use this as the basis for the next set of inpainting, using a different model...

I have a soft spot for big puffy sleeves (and pink). :)

Finally, I choose the best picture and use the upscale tool to scale up the image to a higher resolution. This really helps sharpen the picture and improves the final quality. Et voila...

Hope that's useful for any other aspiring AI image makers?

Feeling nice today, so wanted to post a selfie here 💜💚💜💚💜