More mermaids
   
   
   

k-popping
   
   
   
   

stills from a preggo war film that doesn't exist:
   
   
   
   
   

same but scifi theme
   
   
   

Baristas
   
   
   
   
   

Nice that we now have a subsection for AI creations, and the cartoon section has been renamed to the Artwork section, makes sense. 

One last picture post from me in this thread. Today it's Nuns.

What would people prefer, should I just post my content in my own thread, or have a thread for each topic/theme? I'm leaning towards the latter.

   
   
   
   
   
   
   
   
   

Separate themes/topics would be great @"AIEOU", love your stuff!

If I get a theme/topic going I will make a thread for it, random stuff can go in here. I have rushed out a few substandard ones today. They have potential, they could do with more work but they aren't terrible:

"two beautiful young women winning a sexy bikini carnival, large breast, bare pregnant bellybutton"
       

Another thing I do is rework lower resolution creations. The final prompt and other settings may be different to the original and difficult to recreate without precise documentation of the steps taken, but it's more productive for me to follow this process rather than just using a prompt to generate images all day and hoping for some winners. I get better results with the faces doing this rather than using upscaling with AI face repair, sometimes other things work out well, sometimes the whole thing gets ruined. First prompts count more for general shape and setting, second prompt is better for description of details. Generally the simpler you can make the prompt the less weird crap happens. 

You can add, remove, edit and rearrange prompts as seems right at the time. Eventually I will work on manual retouching to improve and reduce blemishes on the belly, a lot of almost good outputs not shown were ruined by just too much weirdness or too many fingers. Some look great as a thumbnail and then a total monster of the uncanny valley in full view.

One rule to remember with prompts is that if it's stupid but it works, it isn't stupid. The effect of words can be indirect but bring better results than direct description.

"pregnancy, two gorgeous curvy young friends winning a sexy competition, large breast, bare pregnant bellybutton, cute adorable makeup, magic garden"
               

I've started a DeviantArt to share my creations here, not looking to monetize it just seems like an appropriate platform. There's some great quality work by other AI creators on there as well. 

Not many posts yet I don't have much time atm, this tech is quick to create okay stuff but it still takes a while to polish things up nice.

   
   
   

   
Playing with merged models (SD 1.5 and NAI, biased towards the former)

Starting to hit some limits of what I can do with free online services - the free google drive doesn't have enough room to play with model merging and free colab doesn't give enough time to get everything done. Looking forward to getting a PC that can handle SD.