My first (and probably last) attempt at AI qs

[cerberus]

I'm not sure if this is going to work with the tool you are using but the prompt below gave me the attached image:

woman with long dark hair and arms raised wearing red shorts and long sleeved white shirt in a mediaeval bedroom sinks deep into wet quicksand floor in front of a large brown bed

The negative prompt was:

(worst quality, low quality, illustration, 3d, 2d,low quality, jpeg artefacts, blurry, poorly drawn, ugly, worst quality), open mouth, tooth, ugly face

This may be similar to what you've tried already, but this, in conjunction with a seed image, may help. Another thing is that there may be a couple of sliders, with SD they're called CFG and Denoise Strength, moving these to a lower setting can constrain the AI to stick closer to the prompt rather than do it's own thing.

00917-3450752445.png

Note: I had to do a few generates to get this, it kept having her kneel down.

There is a free option that should do what you want, however there is a big "but". This is the tool that I use, it's called Stable Diffusion and can be downloaded with a choice of interfaces (the one I use is Automatic1111, there are others). The "but" is that it requires a decent spec machine with, ideally a NVidia good spec graphics card with a minimum of 8GB graphics memory, mine has 12GB and I often exceed that (most of the time in fact). The initial install disk usage is somewhere around the 30GB mark, but it can grow rapidly. The other part of the "but" is that there is a lot to download and install, it is not a simple one click windows setup and there is a steep learning curve. If this option interest you I can suggest Youtube vids that go through the process and advise on extensions, checkpoint models etc. but this route does involve a significant commitment in terms of time.

[Viridian]

Actually, Viridan, I was using my single-frame update pic "Christmas Magick 2024" that I posted in January. No fountains, just a bed and a qs carpet. lol
But I like what you did there! Especially evil4.png. Is there no way to achieve that sort of result using just one AI program? And do it for free? I'm not ready to dump any money into it, at this point.

Oh, in that case:

Magick Comic.jpg

Magick Photo.jpg

Can it be one for free? As cerberus covered, you can download Stable Diffusion and run it off your own computer, but this requires no small amount of hardware and time investment to understand what is essentially the back end of AI art. The apps that we use greatly simplify that for us by giving us easier interfaces and cloud computing power, though the best models and features may be behind a paywall.

The challenge with creating this kind of image, in addition, is that the model might not be trained to create the kind of image you want. This is why I've said a bit about the kind of 3D model poses you tend to use - for an AI, the outfits are unconventional, the poses are confusing,and the setting is complicated. This makes it hard to generate your images with photorealistic models, because they're not trained to create things like quicksand rugs and floorboards, so when we use a source image, the AI will heavily lean towards trying to make sense of the scene with your prompt. An example of the lack of clarity is your magic mist effect, which is more likely to prompt the AI to think it's water.

It's much easier to create this with my comic art prompts, partially because there is far more data that is used to generate the art style and it is not so inflexible when creating this unconventional scenes and outfits.