Quicksand Fans

Since virtually all posts here are all improbably large breasted girls, this is probably only for the minority here who prefer guys.

These were done with Google's AT Test Kitchen ImageFX which will generate images in your browser. It does require a few tries to get the image you might want (crafting the text prompt can take some trial and error, the depth of the mud seems tricky to get right) and there's some kind of limit on how many images you can create a day so you can quickly burn through these by adjusting your text prompt, but it'll often produce some pretty good results.

Welcome to our little group. Admittedly these aren't really my type, but still, nicely done. I haven't come across any Google AT ... ones before so Interesting to see how they worked out. All the AI generators seem to need a few tries and what you say about the prompts also seems to apply generally as well. Anyway, keep up the good work.

This is what I've waiting for, thank you for it!

and for this topic, I just generated some male video game characters in real life and sank into quicksand.
but i can't show image tho...

A few more, with a bit of tinkering with the prompt you can alter the camera angle a bit. It's too bad there's no way to keep certain parts of the composition, e.g. the guy it generated, and change things around.

It's all perfect, how about making a man sinking in quicksand story using AI? it would be great.

At least with the AI test kitchen, storyboarding an image story won't work just yet, there just isn't the controls to make it generate the same character in each image (the only control you have is of the seed, using the same seed with the same prompt will generate the same image, but with a different prompt will do something completely different).

Anyhow here's some in a bit of a different style to the last ones.

Great pics - more men in mud please!

Love the pics! Great start! Keep on going, would love to learn how to do this.

I've been experimenting a bit with Kling (I didn't initially want to bother as it only works if you have a gmail account or similar, I discovered I had a long unused gmail account to sign up with). An advantage it has over Google's ImageFX is that you can provide an image to work off, and so you can make many images in many situations with the same person. There are some limitations, for example, I could never get Kling AI to make an image where the man was facing a different direction, the images it generated were always with the person facing the same way as they were in the original image regardless of what you put in the prompt.

However I got some decent results using ChatGPT's DALL-E to recreate an image where the person is facing in a different direction, then in turn use that with Kling (when I get a bit of time to sort the images I will probably post some of these)

Kling doesn't seem to be quite as photo real as ImageFX, there seems to be a slight softness to the images, but the results are pretty good. For instance:





I also took one of the images I made with Google ImageFX and used Kling to modify it (the original picture is upthread, the rather good looking guy in the white singlet looking up at the camera)



One thing I've really found hard to do with Kling is generate mud that isn't watery (particularly for video, neither Kling nor Flux seem to be able to generate mud that doesn't just turn to brown water), and also have the guy sinking in it much more than about waist deep, it took many, many attempts to get the first picture. ChatGPT is better in this respect as you can ask it to modify the picture more and it remembers what it last did, but ChatGPT doesn't do the mud nearly as well as Kling.

I think at the moment my ideal AI would have: the mud, background and character generation of Google ImageFX, the consistency of Kling, and the building on previous ability of ChatGPT.