Renderotica

Originally Posted by: Finister

Poser and DAZ had a learning curve, but with AI someone just needs to add text inputs. the text inputs have their own creative uniqueness perhaps.

You hear people _say_ that -- but if you want to do anything at all interesting, get two figures doing something to each other for example, particularly something sexual -- that's hours of work.

The technology is now evolving incredibly quickly, and the biggest new thing is "ControlNet", which basically allows you to pose "figures" (although figures don't really exist in Stable Diffusion) to guide the models to the kind of pose you want.

Its a different way of doing things to Daz or Blender, with strengths and weaknesses.

Like both of them, the uninspired and the lazy will load up something that they got from somewhere else, click "render" and get something mediocre. Getting something more than mediocre, that requires a lot of wrangling.

For Poser veterans, particularly those vendors who own and sell valuable assets -- its time to at least consider looking at how to sell variations of those assets into the Stable Diffusion community. There are lots of folks selling "prompts" for a lot of money, also folks selling models (which are not 3D models, more a statitiscal model of imagery).  There's money to be made there, and there's money to be made in precisely the categories of imagery that Renderotica vendors have been selling -- but its going to need getting acquainted with the technology and figuring out what and how to sell.

Personally I'm not bothered about it, just as long as they are labeled as such, which they seem to be at the moment, with a separate category, which there is. For me, it will only become an issue if someone tries to pass it off as Poser or DS, or their own work....Relying on people's honesty isn't always rewarded...

I think all of this is hilarious.

Here you are, thinking that you've come up with someting new and yet... had you bothered to dive into this thing more thoroughly...

It's nothing new!

(edit): Search youtube for: "reason a-list guitarist".

^ Here's "AI" playing an accoustic guitar 8 years ago.

(edit): Search youtube for "ni electric sunburst".

^ There's "Electric Sunburst", electic guitar AI playing 5 years ago.

awwww, crap... external URL's are a no no here, it's all youtube but... I respect them rules.

Thing is: we've been here before, digital audio is way more static than people think. Seriously: the hits of this century? I hear familiar 808's, 909's and familiar percussion patterns. Some things don't change. And yet...  ^ see above. 10 years ago we already had efforts to automate background percussion, background jams and .. even more. HECK, Synths themselves should have taught you that...   NOPE.

AI is overhyped, just like it has been 10 years prior within the field of digital audio.

Not to mention... ChatGTP?  It's not that much different from chatterbots back in the 90's. Seriously...   it responds to keywords and in return presents you with a text. Been there, done that. The main difference: bigger databases with stuff to respond to.

No personal offense intended but you're delusional if you think that this is going to change anything. HECK, men and women still have differences to understand each other sometimes, and unless you manage to solve THAT no AI is gonna fix anything.

Originally Posted by: davo

I think those 2 artists set the standard for modern fantasy art. If anybody remembers NVIATWAS that was what those 3d enthusiasts were trying to do. I really do wish this site tried to embrace more artistic style artwork rather than photo real, that's where the artist comes out IMHO.

I've started taking a look at Vallejo and Frazetta. The short form is -- its going to be hard. The thing of it is -- they're good artists, and they used their talents with a lot of discretion, even though it looks very over-the-top, they actually modulate tone. I generated a pile of these to with varying prompts and feedbacks -- none of them rate as finished product, obviously - but this gives you an idea of how Stable Diffusion works in an artists hands. its not "magical prompt and now you have art" -- its "get some pieces, and then sttch them in Photoshop, and then send them back to Stable Diffusion with new iterated prompts" --  that's how you get some interesting things.

But since folks may be interested in seeing what "the raw material" is like coming from some of these prompts, here is it is. One of the thing I noticed was that nearly any mention of men and muscles gives rise to a lot of homoerotic material - nothing wrong with that, but if you're going for, say, Conan - you're skating this very fine edge . . . Frazetta and Vallejo merge into Tom of Finland really easily. You can also see that other than hints of particular images, the style isn't really all that distinct from other fantasty artists, Hildebrandts, Barry Windsor Smith and so on. So that's just one of the many reasons why the folks who're trashing AI art don't know what they're talking about -- to get something good, you have to really understand what art is about. It ain't just click a button "make art" - there is a button, but if you don't do anything more with it, what you get is pretty much the same as image that just got clicked. Like loading V4 into Poser and clicking "render"

All of these  have major things wrong -stylistically, anatomically and compositionally. But they also all have some things that are right . . . so that's how you do AI art, the interesting kind anyway. Generate ideas, stick them in your sketchbook, think about how they might work in a composition, rework the stuff that doesn't work and so on. So I think we'll see these again, but I have to figure out what to do with them first

Originally Posted by: Darkseal

I've put in the work these last 9 month with ai. I did a commission for a client that took 4 days to make. The work has shifted for content creation, that is all. I'm not going to go into any detail about my thoughts, I dont think I'll change anyone's mind here. As a content creator for the markeplaces for some time, my main hope is that, since it and the derivitives of, are all 100% legal and, depending on license can be commercial free, My main hope is that places that I've been making content for over a decade could find a way to open an ai marketplace. This is the result of some work I've done for a client. Not the girls, the ball gag harness. These renders take me about 10 seconds on 8gb 2070. I would never get these results in octane, the plugin I use for poser. You feel me?

No argument from me. I'm all in on AI.  I'm still using Daz to do things like pose hands, and Blender to model some things . .. but this is a tidal wave, and the incredible thing is that the rate of change is increasing, just look at the Midjourney 5 developments today. Sadly Midjourney ain't friendly to NSFW, but you're already seeing Midjourney like models in Stable Diffusion.

Originally Posted by: Finister

What do you all think about AI art?

I did not think this fully through yet, but my main concern is, that with this little effort AI will produce such numbers of "art" that the single piece becomes meaningless.

You had this already with the step from analog to digital photography. The whole internet is flooded and the really good stuff ist hard to find.

In some cases artist are succsessful to make a brand on their person, but this doesnt mean quality.

Originally Posted by: topspin

Rest assured -- it won't

Most AI artists -- can't do anything but "render pretty girl"

You can spew out a zillion "render pretty girl" - and folks get pretty bored of that.

In my own work that I post, I no longer do any AI that doesn't have an interesting setting, characters interacting with each other in an interesting way, or an aesthetically interesting stylization. And now I'm trying to all those things in one image.

So now I'm working on the current "Alien Invasion Challenge"  . .. but that's like, work.

That‘s the right attitude! I am curious what you come up with in the contest to see what AI is capable of at its best.

Hopefully someone will develop also an AI algorithm to clear the million "pretty girls" from our screens.

Originally Posted by: slack71

So, while I do consider AI images a form of art and have seen some very interesting pieces, I do not think it should be on a site for 3d art even though it can create art that imitates 3d, it really is not. They should not be allowed for the contests and if AI is continued to be allowed here, there should be someway to separate it completely from actual 3d rendered images. Just because one thing looks like another does not mean it is actually the real thing. AI Art is going to destroy this site as I have already begun coming here less and less as the AI art grows. I am sure I am not the only one. I have noticed not as many new images are getting posted to the gallery and most of the ones that do get posted are AI Art. Not even going to enter the Alien contest because I see obvious AI art already entered.

What I'd say is look at little harder at the subject.

There's a lot of bad, lazy AI, just as there's a lot of bad, lazy 3D art. Lots of folks can't be troubled to even try to light a scene, for example . . . this is years after HDRI and iRay revolutionized the possibilities for Daz/Poser art.  Much of what I see produced by Daz/Poser artists looks like the crude renderings you might have had of V4, almost 20 years ago.

. . . but of course, other folks invest huge energy in posing, lighting, materials -- and they come up with something completely original.

AI is like that . . . I suppose with the exception that some some entering a prompt "hottest girl ever, 4K, 8K, model, movie star, big tits, photorealistic" -- they'll get a competent boring photograph of the subject.

. . . but in general, bad lazy work without any idea -- pretty much is all boring.

AI is now implemented as plugins in Blender and Photoshop, and I imagine it will be in Daz too, I would bet on "any minute now"

It can be used in all sorts of ways. Daz users can use it simply to style an image, or to alter it slightly. As an example, here's the one remaining Daz image that I have posted here, and then some variations I generated on it from Stable Diffusion. You'll see the opportunity here -- I had an image that I kinda liked, but which had problems- stiff arms, no hair (though that was the idea at the time).

- that was %100 Daz, not really very impressive - a model, a set, about the only real creativity there is an unusual camera angle and nice lighting, but I never got the hair the way I wanted it, so I left it off.

Now I can generate a zillion variations on that image, fuel for tweaks, changes to styles and materials

. . . so that's just one way that Stable Diffusion can improve the output from your Daz images. It can, of course, do much much more. [None of these are things that I'd post -- all have little gotcha that need fixing - but look at how good it is in generating a consistent lighting and texture scenario across a scene, and how well the hair gets added.

. . . or you can produce a zillion crappy renders of a poorly posed, badly lit and textured model.

So it all really comes down to having a critical eye for your content - "what's good enough to be worth showing to someone else"?

Stable Diffusion rendering has already been implemented in Blender as a plugin, and I imagine it will arrive in Daz very soon . . .

So its not "3D or AI" - its "how do you use these tools to make interesting images"

- just as folks buy $3000 cameras and produce crap photos . . . so too these tools, its about what you do with the possibilities as much as the technology. . .

 . . . this variation, which rendered it as a watercolor and changed a lof ot materials and lighting, I liked. You _can_ do non photorealistic rendering in iRay or Daz, of course, but Stable Diffusion route is so much faster, offers stylisitc consistency, generates lots of variations to consider.

. . . and here are two more versions, where I "loosened" the parameters (CFG and denoising) and "pushed" Stable Diffusion towards a different angle. Its much easier to quickly iterate through options like this in SD, vs Daz - but the essentials of the pose from Daz still remain, even if a lot has changed

Edited by user Tuesday, March 21, 2023 5:00:02 PM(UTC)  | Reason: Not specified