Stable Diffusion AI Image Generator

I've just realised that we might not even be using the same AI models.

The setup I followed a few months back had multiple models.

I assume you're all using the original stable diffusion v1.4.
 
I've just realised that we might not even be using the same AI models.

The setup I followed a few months back had multiple models.

I assume you're all using the original stable diffusion v1.4.
I'm using 1.5 and a bunch of other models for different styles. There's also a special model for improved inpainting and outpainting.
 
I used IMG2IMG on a comic book drawing of Superman to make it photo realistic and then used it again to deepfake Henry Cavill's face in. Then some minor tweaks in photoshop and it's like the character stepped right off the page of a comic book into real life... all in less than an hour.

Comicbook Superman.jpg
 
I too have been messing around with this for some of my beer labels and have been loving it.

Currently using https://creator.nightcafe.studio

Love the fact that I dont need to use anyone elses art and can have something ultra unique!


Irish Red Ale

1668410079484.png

1668410088823.png

English special bitter

1668410655041.png

Sometimes it pure nightmare fuel though

1668410721532.png

Octopus in a hops field

1668410689702.png
 
Some security related tutorials if you're running a local SD and have downloaded a few models which might contain malicious code.


 
I too have been messing around with this for some of my beer labels and have been loving it.

Currently using https://creator.nightcafe.studio

Love the fact that I dont need to use anyone elses art and can have something ultra unique!


Irish Red Ale

View attachment 1421295

View attachment 1421297

English special bitter

View attachment 1421303

Sometimes it pure nightmare fuel though

View attachment 1421307

Octopus in a hops field

View attachment 1421305
That top one is a keeper.
 
I don't mind if we mix up sharing renders from SD and tutorials \ technical stuff for SD here? The DALL-E2 thread kind of turned into a general AI image generator thread, so I suppose we can use that for anything that isn't SD related.
 
As in results only from SD? My posts so far are SD results. Or do you want to keep it SD technical related?
Results from SD are also cool here imo. Maybe need another thread for general AI images.
 
Gothic Vampire woman, upscaled with R-ESRGANx4, then upscaled again with Topaz Gigapixel (and then downscaled to post here :laugh:)

Got some very impressive detail in this pic.

Vamp.jpg
 
Damn this will never run on my Dell 5290 2-in-1 :(
There are many ways to run it if you don't have a high spec GPU or PC... Google Colab, Runpod, etc. Or you could go with one of the many web services, some are free AFAIK.
 
I discovered some lesser known tidbits about SD recently. It turns out you can blend the features of multiple people to essentially create new characters or to average a particular character that has been portrayed by various actors.

In this example I blended together all the actors that have played the role of James Bond...Sean Connery, George Lazenby, Roger Moore, Timothy Dalton, Pierce Brosnan and Daniel Craig even though I blended all six equally, the result mostly resembles Brosnan with some Connery and Craig's bright blue eyes in most renders. Maybe I just don't know Lazenby and Moore well enough to recognise which features are theirs.

James.jpg

James2.jpg

And... even what's even more interesting is you can even render this blended character at different ages... here's a younger version of James...

Young James.jpg


Powerful stuff!
 
png.png

Deforum. Used a video of Pong^^^ (the old paddle game) as an input and got out a frantic series of disjointed frames. Guess the black background added to the way the images were made. Not that interesting to watch but some decent images in there nonetheless. Started editing it with some music but lost interest as it's just too much to manually sift through.

 
Top
Sign up to the MyBroadband newsletter
X