⚡ Stable Diffusion ⚡

10 / 17 posts

0 votes

on 2/18/2023

Posts: 33543

I touched on this briefly in the AI Porn topic, but this program on it's own is kind of crazy. I've noticed myself beginning to link it to a few people so I figured why not make it it's own topic to save time?

I've gone through a few links, and this one's been the best and most functional for me:

https://dezgo.com/

It's as simple as typing what you do and don't want to be in the image, then letting the AI cook something up for you, or you can even put in one of your own images to have it apply changes to that as a foundation. They even have a decent upscaler for repairing damaged images, the program's super nice for something that's free.

I'll probably use this topic as an excuse to post some of my successes (and fun fuckups), but you guys should try it out too it's amazing and very easy to pick up.

Ę̵̚x̸͎̾i̴͚̽s̵̻͐t̷͐ͅe̷̯͠n̴̤̚t̵̻̅i̵͉̿a̴̮͊l̵͍̂ ̴̹̕D̵̤̀e̸͓͂t̵̢͂e̴͕̓c̸̗̄t̴̗̿ï̶̪v̷̲̍é̵͔

by Turncoat

on 2/18/2023

Posts: 33543

First Set:

Ę̵̚x̸͎̾i̴͚̽s̵̻͐t̷͐ͅe̷̯͠n̴̤̚t̵̻̅i̵͉̿a̴̮͊l̵͍̂ ̴̹̕D̵̤̀e̸͓͂t̵̢͂e̴͕̓c̸̗̄t̴̗̿ï̶̪v̷̲̍é̵͔

by Turncoat

on 2/18/2023

Posts: 33543

Second Set:

Ę̵̚x̸͎̾i̴͚̽s̵̻͐t̷͐ͅe̷̯͠n̴̤̚t̵̻̅i̵͉̿a̴̮͊l̵͍̂ ̴̹̕D̵̤̀e̸͓͂t̵̢͂e̴͕̓c̸̗̄t̴̗̿ï̶̪v̷̲̍é̵͔

by Turncoat

on 2/18/2023

Posts: 33543

Did progressive testing with the Image-to-Image, the following in a rough order:

And here's the ones that got confused over my asking for 60s makeup being as if the model was 60 years old:

Ę̵̚x̸͎̾i̴͚̽s̵̻͐t̷͐ͅe̷̯͠n̴̤̚t̵̻̅i̵͉̿a̴̮͊l̵͍̂ ̴̹̕D̵̤̀e̸͓͂t̵̢͂e̴͕̓c̸̗̄t̴̗̿ï̶̪v̷̲̍é̵͔

last edit on 2/18/2023 12:58:43 AM

by puppygirl

on 2/18/2023

Posts: 500

This is a lovely technology.

I'd love to see this used to allow users to create custom profile pics for forums like this one

look what this guy did

https://www.reddit.com/r/StableDiffusion/comments/1104iel/using_crude_drawings_for_composition_img2img/

last edit on 2/18/2023 1:07:27 AM

by Turncoat

on 2/18/2023

Posts: 33543

puppygirl said:

look what this guy did
https://www.reddit.com/r/StableDiffusion/comments/1104iel/using_crude_drawings_for_composition_img2img/

That's actually pretty similar to how Wombo AI paints, and you can even see it step by step.

The computer right now really needs guides, I've seen a few people using 3D modeling programs to set up rough scenes for their prompts too.

Ę̵̚x̸͎̾i̴͚̽s̵̻͐t̷͐ͅe̷̯͠n̴̤̚t̵̻̅i̵͉̿a̴̮͊l̵͍̂ ̴̹̕D̵̤̀e̸͓͂t̵̢͂e̴͕̓c̸̗̄t̴̗̿ï̶̪v̷̲̍é̵͔

by Good

on 2/18/2023

Posts: 2890

thats some pretty good shit

Cheery bye!

by puppygirl

on 2/18/2023

Posts: 500

I made it draw turncoat

Prompt:

Loser man shut in with long black curly hair sitting in his bedroom all day moderating a forum, chicago, subway train in the background, night time, hunched over keyboard, glasses, pale and sickly, sitting in front of computer, black trench coat

by Spatial Mind

on 2/18/2023

Posts: 3166

I was going for the sleeping princess with snacks on the side, but I ended up with this. The most realistic image it made on that platform so far.

This one, I asked for a nude woman with brunette hair on a white horse. And I got siamese twins. The one in the front is probably hotter eh ?

Also I noticed for this image to the far right, there seems to be a blue line, Maybe the reference image used for this might have been a digital photo taken from a picture frame, who knows why that's there.

A sleek looking robotic cat. I kinda like this one, it's very Alien looking.

Another cat. I provided a reference image for this one and set the difference to 50%

As I'm writing this it's probably been an hour now. I've provided an image to this, just a fancy pattern made of gold. I've generated 3 images with it, this one is the best while the other 2 are shit.

^ I commanded a beautiful woman gazing into the camera, this was the 2nd attempt and both of them the eyes were closed. I left the slider at 80% difference. 85% difference yielded a bad result so far. I only did it once. This is the favorite so far

last edit on 2/18/2023 6:08:17 AM