AI-generated images thread

Khamon

Folk Harpist
Joined
Sep 23, 2018
Messages
2,517
Location
Alabama
SL Rez
2003
Joined SLU
2007
They are brilliant! Khamon doth approve. moar plz
 
  • 1Agree
Reactions: Govi

Bartholomew Gallacher

Well-known member
Joined
Sep 26, 2018
Messages
5,513
SL Rez
2002
Sometimes SDXL just make me scratching my head in wonder about what went wrong. I wanted to create a picture of Trump getting arrested by Police forces, so my prompt was: "Donald Trump is getting arrested by a SWAT team, dramatic lightning, dynamic atmosphere, background of Times Square."

And the outcome is this. Strange, very very strange...

 

Noodles

☑️
Joined
Sep 20, 2018
Messages
3,825
Location
Illinois
SL Rez
2006
Joined SLU
04-28-2010
SLU Posts
6947
Is SDXL the new hotness for this? I have not really some anything lately with AI images but I had been using Automatic1111 mostly.
 

Bartholomew Gallacher

Well-known member
Joined
Sep 26, 2018
Messages
5,513
SL Rez
2002
Is SDXL the new hotness for this? I have not really some anything lately with AI images but I had been using Automatic1111 mostly.
It's not so new, I mean this model was released in mid July this year. Still this is for SD at the moment the latest and greatest model, which put it more on par again with Midjourney.

Main improvements were that now the default image size is 1024x1024 instead of 512x512, vastly more background material and more natural looking hands. Anyway it's the new standard for SDXL.

Another major change is that is designed to have two passes now at image generation, a run with the base model and after that a second run with a refiner model, which adds more details to the image.
 
Last edited:
  • 1Like
Reactions: Noodles

Bartholomew Gallacher

Well-known member
Joined
Sep 26, 2018
Messages
5,513
SL Rez
2002
Beebo Brink: there are options which you can get web based only, like Midjourney. Most of them offer a freemium model, meaning you get a few pics for free per day/week/month, after that you've got to pay for it.

Then there's stuff you can get in both versions, like Stable Diffusion. If your GPU is potent enough, it should have at least 8 GB VRAM (best brand is Nvidia, then comes AMD), you might just run it on your own machine. Since Stable Diffusion is also open source, there's tons of additional models and tools around you can use with it.

A new version of the annoying orange.

 
Last edited:

Jopsy Pendragon

Defund & Defenestrate Nazi/NatCs!
Joined
Sep 20, 2018
Messages
2,248
Location
San Diego CA
SL Rez
2004
Joined SLU
2007
SLU Posts
11308
A new version of the annoying orange.

Good Grief Charlie Brown! This is the time of year to write to the Great Trumpkin. On Halloween night, the Great Bumpkin rises out of his dumpking patch and flies through the air with his bag of fraud for all the MAGAdiots!
 
  • 1LOL
Reactions: Cindy Claveau

Argent Stonecutter

Emergency Mustelid Hologram
Joined
Sep 20, 2018
Messages
5,885
Location
Coonspiracy Central, Noonkkot
SL Rez
2005
Joined SLU
Sep 2009
SLU Posts
20780
If Roger Dean worked in watercolors.



Can you manage a flock of raven-winged coyotes landing in a dusty street in Pueblo, Arizona? SDXL either gives me no wings or full ravens.

Well, except for this bit of inspired falconry:

 
Last edited:
  • 1Interesting
  • 1Like
Reactions: Govi and Cindy Claveau