Viridian's AI Experiments

[sixgunzloaded]

Definitely all beautiful! Great variations on these characters. I like how the one in "David's Day" is grabbing the qs like - 'What the hell is this?'
If you're feeling bored or uninspired with AI stuff at the moment, why not focus on some of your drawn art instead? That's always nice too!

[Viridian]

Definitely all beautiful! Great variations on these characters. I like how the one in "David's Day" is grabbing the qs like - 'What the hell is this?'
If you're feeling bored or uninspired with AI stuff at the moment, why not focus on some of your drawn art instead? That's always nice too!

That means not spending time refining the AI craft.

[sixgunzloaded]

Definitely all beautiful! Great variations on these characters. I like how the one in "David's Day" is grabbing the qs like - 'What the hell is this?'
If you're feeling bored or uninspired with AI stuff at the moment, why not focus on some of your drawn art instead? That's always nice too!

That means not spending time refining the AI craft.

True, of course. I just mean sometimes it can help to take a break and then come back with fresh eyes/thoughts.
Also, more awesome pics, by the way!

[Viridian]

So there's been another turning point. It's been a dream to literally type in something like "cheerleaders in quicksand" and get exactly what you want. That function is now available with Dall.E 3 and ChatGPT. The results are coherent, though the quality of output leaves something to be desired. Running it through a Stable Diffusion model can get nice results, but the tangle of faces and limbs makes it very confusing for the AI.

I also tried a scene from "Gail's Gallop", which I could never generate with AI alone. ChatGPT was able to create the scene, which I then rendered into SD.

[Viridian]

Further updates. The use of ChatGPT with Dall.E 3 has really opened up the complexity of generations. As other models can't interpret design elements specifically and leave it up to chance, ChatGPT is far better are shaping what you want, especially when asked to change or refine things in plain in English. As a result, I've been able to create the source images for scenes for artwork that I had previous been unable to render, and run those through Stable Diffusion.

[Jinn]

The cheerleaders… Holy, all those helpless girls sinking at once… My brain is having a hard time processing… @.@

[cerberus]

A bus load of cheerleaders in a swamp, what's not to like. The others are cool too. The comparison between the Dall.E 3 and SD versions is very interesting, thank you for posting these, they are thought provoking. My experiences with Perchance have been a bit frustrating, in that it is difficult get it to do what you want and you just have to do lots and pick the closest. I will be following developments here with considerable interest.

[dunebar]

Great work on the very grainy dry sand texture, that's been something I've struggled to get consistently out of a prompt despite using every "grainy" related word I can think of, in a hundred variations, haha.

[Viridian]

Another set of costumes.

[sixgunzloaded]

I love these big collections you do. So much to please the eye, yet so hard to pick a favorite!
The firefighters are really cool. The encroaching fire in the background makes them really eye-catching. The ladies better hurry up and get out of there!
Or... on second thought, maybe they're better off right where they are in the nice wet mud that won't burn!

And those cheerleader shots are incredible! All that flailing, splashing mud is fantastic! And the looks of absolute terror on all their faces are the most intense yet!
I like how the quicksand is all humped up all around them, as if it's alive and actively consuming them. No wonder they're in horror! Nicely done!