Stable Diffusion and Fansadox

[xyzzs]

Has anyone else experimented with this?

Using stable diffusion and controlnet.

[xyzzs]

[Cossack]

I've started playing with this too after being inspired by a thread on folks doing this with Erenisch characters. Would love to swap notes if you're on Discord or something?

[skarab]

-

[xyzzs]

It may seem daunting at first but it's actually quite simple if you have a decent Nvidia card.

The basic process is:
-install stable diffusion
-grab a good model from https://civitai.com/ (Deliberate is a good one to start with)
-put the downloaded model in the appropriate folder
-install controlnet
-use canny/openpose in controlnet to replicate the image you want to use as inspiration

This video shows you how to install stable diffusion easily
https://www.youtube.com/watch?v=3cvP7yJotUM

and this one shows you how to install controlnet and set it up
https://www.youtube.com/watch?v=OxFcIv8Gq8o

I'd recommend just installing SD and a model and playing around with it for a bit before you try and use controlnet. Prompts for each model can all be found on the images uploaded to Civate or by searching for "stable diffusion prompts".

I went from knowing nothing to generating what you see above in about a week of just playing around. There a ton of tutorials and information out there talking about this stuff and it's only getting better every day.

Some good places to get info on stable diffusion are
www.reddit.com/r/stablediffusion
https://rentry.org/sdg-link

It will seem overwhelming at first but once you play around with it a bit, you'll get the hang of it and it is very addictive.

I'm a novice myself but I'm happy to help out anyone as far as my knowledge goes. You can also use AUTOMATIC1111 (the go-to gui that makes everything easy to use) on Apple silicon. It's not as fast but still useable. I didn't have much success with my AMD card and swapped out a 6600xt for a 3060 12g and it excellent for this sort of thing.

I just ran some quick googling on SD and I learned it’s all about training the AI to recognize image, and it would output the CGI, so I’m like “what great idea” I can just as easily upload my drawn comics panel and it would do the rest, then I thought about the angles and perspective. Would it recognize that too?

With controlnet and SD it's as easy as pointing it to a picture as I did above and it will perfectly pick up the pose. You can even throw together a rough collage in photoshop and it will turn it into a picture. It really is amazing tech, I'd encourage you to check it out.

[xyzzs]

This one from Moffet's "Deadly Platform" has some problems that I didn't fix as it was more of a proof of concept (the hands, face, left nipple etc).