Daz 3D is part of 
Connect
DAZ Productions, Inc.
7533 S Center View Ct #4664
West Jordan, UT 84084
Licensing Agreement | Terms of Service | Privacy Policy | EULA
© 2025 Daz Productions Inc. All Rights Reserved.
Comments
Nice (I suspect she's making a sultry expression under that helmet though*)
*J/K obviously, but I mention it in case it's not obvious.
LOL
(Now, just a quick note before I go on. I'm pretty sure you are joking due to the preponderance of the style of images in question, so the rest of my response is more at those who might not understand that AI really does do more than that.)
I can definitely see where you are coming from since that vast majority of the AI images you see would seem to support that idea. However, it's just like for someone who has never gotten into really using DS (or Poser) it would seem that it is only good for producing NVIATWAS, and in the early years it was really bad quality NVIATWAS!! (Note for those who haven't been around that long, this stands for Naked Vicky In A Tomb With A Sword, this seemed to be the most prominent art form in the galleries back in the Poser 3/4 days.)
Judging what a product can produce simply by the digital "art" most prominently posted on line runs the risk of thinking it is only capable of that art form, when it really isn't limited to only that. So, to counter your tongue in cheek argument that you need to put in extra effort to not get the AI equivalent to NVIATWAS, I will provide the following example.
Using this simple prompt "redhead female in a Victorian gown standing in the foyer of a Victorian mansion" resulted in the image posed below (it was the first image produced using the Krita AI plugin for Krita). I'm not a big proponent of AI image generation, but I am investigating it to better understand it potential uses and capabilities. I'm also interested in how it might be used in conjunction with DS.
Now, just because no discussion should be engaged without invoking kitties. this second image is from Krita AI using the prompt "cute siberian gray kitty in a field of flowers, sitting, close up".
I'm constantly amazed at what some of the people who use/incorporate AI image generation are able to produce. With that in mind, here are some links to some images that definitely "break the mold" (if anyone is interested).
https://civitai.green/images/50629679
https://civitai.green/images/50380475
https://civitai.green/images/50623217
https://www.deviantart.com/editpix2/art/Caitax-1144923026
https://www.daz3d.com/gallery#gallery=aipartial&sort=popular&page=1&image=1357193
https://www.daz3d.com/gallery#gallery=aipartial&sort=popular&page=1&image=1357040
I'm glad I wasn't drinking coffee when I read this!
The prompts of the first three are nice short-stories.. but somehow the pictures don't really manage to show fully what they describe. Looks a bit like a Hollywood movie version of a good book to me...
And about those scantily dressed ladies that show up, sometimes (often?) without being invited by the prompt: seems it depends what AI gets used and how it had been trained, right? A wild AI, feeding on all the interwebz does offer, without being restricted somehow, might be produce other content than one that got served a nice menu of Art.
YES!
That's a great analogy as often the image generated isn't quite the same (as one might interpret) as the text, and often a bit of an abridged version of the prompt. Though sometimes it can go the other way.
Yes, the AI model and checkpoint being used can significantly alter the output of the images (as I understand it, a checkpoint is sort of an extension of the base model with additional training data to influence the base ai model to generate images of a certain style/content better). AI image generation can also be influenced by a Lora. A Lora is sort of an external add on to the ai model that is generated using smaller more focused image sets to guide the ai generation to create more specific image elements. For example, if you wanted to consistently create a specific character in your images, you could create a Lora that would focus on that character and different angles/distances and environments, then the ai image generator will produce images using this character when it is referenced in the prompt.
I'm sure commercial sites will train their models to produce the images their customers want to see/create, and no doubt add a lot more training data than found in a base model like SDXL. There are also free checkpoints based on free models like the SDXL model (open source model) that extend it, and can be used to generate specific types of content easier (at least that is how I understand it).
Ah yes. Daz Studio-known its excessive usage of spicy imagery. As bad of a rap it gets for that, I think any platform that can create visual media has adult content as a huge part of its use. After all, it's the best money maker (unfortunately).
I actually find it hilarious that the scantily ladies can just appear in prompts even unrelated to it it's like..."So here's that castle you ordered and...oh how'd SHE get in there? Well, anyway, image is done!"
What a time to be alive.
^This.
I imagine Krita AI is geared more towards the artistic kind of images that Krita is more famous for (at least the images I see advertised as 'made with Krita', which is an ace programme by itself).
Certain free online ai models may have a propensity to throw out the NVIATWAS style. The more control you have over the checkpoints / Lora, the more you'll get the style of thing you're looking for.
I've been unable to get any of the ones I've tried to follow a prompt closely enough to meet what's in my head, but I'm sure there are people who can.
There's someone on DeviantArt (damnmad660) who displays quite impressive results from enhancing the photorealism of his renders with AI. So I thought I'd give it a shot as well. Photorealism always was and is the aim of my renders. I think it depends on many variables. What are you trying to achieve, what the quality of the source file is, what model checkpoints and prompts you are using and so on.
So I took an admittedly older render of mine and dabbled around with different settings in ComfyUI. After about two hours I had a kind of acceptable result which I put into Photoshop and added it as a layer with 50% opacity on my original file. Results below.
Truth to be told I still prefer the original render. Still I wouldn't bet that I'll come to the same conclusion by mid 2026 or so.
One of my favorite dumb AI tricks is still just converting a render and then going back through and in painting section by section to up the detail and quality. I imagine it scratches the same itch as those people with coloring book apps on their phones where they're just tapping predefined colors into predefined spaces.