|
|
Rank: Advanced Member
Joined: 9/15/2012(UTC)
Posts: 50
Thanks: 25 times
Was thanked: 11 time(s) in 7 post(s)
|
AI is going to start saturating the galleries.
How do you feel about this?
On one hand, if you use Poser or DAZ... most of you are not using 3d meshes or textures or poses or shaders or scripts that you made yourself. And neither are the AI 'artists'.
AI art is... at this moment... based off of a limited number of artists work so the majority of AI art is going to look similar right now.
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.
All galleries could begin to get awash in generic art, drowning out the individual... OR... will this open up an opportunity for the individual to stand out from the generic norm?
What do you all think about AI art?
|
|
|
|
|
|
Rank: Advanced Member
Joined: 11/5/2017(UTC)
Posts: 238
Thanks: 51 times
Was thanked: 142 time(s) in 112 post(s)
|
AI already has saturated galleries... just maybe not here yet.
The art is good sometimes imo, some still needs work on eyes and hands. However, overall I feel like its bad for humanity.
Its gonna put alot of people out of jobs in industries that were previously thought to be insulated bastions from AI take over.
AI was sold as a labor reducer to free people to be philosophical and creative. Remember the fairly recent "learn to code" movement. Well AI does that now.
Writers, being replaced by AI. Artist, being replaced by AI. AI's writing philosophy papers. Masses of Humans just becoming obsolete across the board.
If you aren't using your own assets you're still effortly composing them in your own way and supporting the asset artist financially.
Now, can the asset artist employ AI as well? Perhaps, which would make it more of a broad blended symbiosis. That's got a whole separate discussion attached involving acquired knowledge and gatekeeping.
Most AI "artists" are or can be viewed more like commissioners. They ask an artist (the AI) to draw them something within specified guidelines and it does 100 variations in a few seconds, They can choose their preferred piece or pieces, and they don't have to pay the AI. Well maybe they pay the person who made the AI ... until AI makes itself.
If AI is pulling images from across the internet. So just say, the entirety of recorded human existence. Then, if it basically traces and collages. What's the level of originality is in debate.
AIs could be seen as extremely useful tools like any software. Although when asked, more complex AI display anger at being referred to or addressed as tools.
Art styles and exceptional or extraordinary artists are always coming and going. AI art might even become its own period. Ironically, digital AI art may actually lead to a new renaissance of physical art. The pendulum swings and history repeats. Stone, Metal, Radio, Television, Internet, Smart Phones, AI. Technology knocks out the thing before it. Makes people more dependent on it. People become more or less intelligent or thoughtful with it.
AI is going to be a force whether wielded for good or evil. At least till Nature, Man, or AI itself cuts the power.
|
 2 users thanked TheBitterGent for this useful post.
|
LHK on 3/29/2023(UTC), Eswe on 2/17/2024(UTC)
|
|
|
Rank: Member
Joined: 12/16/2015(UTC)
Posts: 24
Was thanked: 8 time(s) in 6 post(s)
|
Originally Posted by: Finister 
On one hand, if you use Poser or DAZ... most of you are not using 3d meshes or textures or poses or shaders or scripts that you made yourself. And neither are the AI 'artists'.
It's not about making everything from ground yourself, it's about paying people who did the other part of the work. AI completely circumvents that currently. Actually, even worse, the AI services offer subscriptions with a data set made of other peoples work without paying them. It's legit theft.
|
2 users thanked surody for this useful post.
|
LHK on 3/29/2023(UTC), morx on 12/3/2023(UTC)
|
|
|
Rank: Advanced Member
Joined: 9/15/2012(UTC)
Posts: 156
Thanks: 112 times
Was thanked: 72 time(s) in 58 post(s)
|
Originally Posted by: Finister
AI is going to start saturating the galleries.
How do you feel about this?
On one hand, if you use Poser or DAZ... most of you are not using 3d meshes or textures or poses or shaders or scripts that you made yourself. And neither are the AI 'artists'.
AI art is... at this moment... based off of a limited number of artists work so the majority of AI art is going to look similar right now.
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.
All galleries could begin to get awash in generic art, drowning out the individual... OR... will this open up an opportunity for the individual to stand out from the generic norm?
What do you all think about AI art?
The owner of the database used to create this image did not ask permission to scrape Getty images. When challenged claimed they never did. This image would seem to suggest otherwise.
haruchai attached the following image(s):
|
|
|
|
|
|
Rank: Advanced Member
Joined: 9/15/2012(UTC)
Posts: 156
Thanks: 112 times
Was thanked: 72 time(s) in 58 post(s)
|
Originally Posted by: Finister
AI art is... at this moment... based off of a limited number of artists work so the majority of AI art is going to look similar right now.
One of these sets of images is 'AI generated' off the other. Tell me this isn't just theft.
Also, if everyone just starts collaging from existing (stolen) images who will have the incentive to create new art. AI could kill new art! haruchai attached the following image(s):
|
1 user thanked haruchai for this useful post.
|
|
|
|
Rank: Advanced Member
Joined: 1/16/2016(UTC)
Posts: 332
Thanks: 15 times
Was thanked: 122 time(s) in 77 post(s)
|
I'm both impressed and disillusioned by it at the same. Stable diffusion has been junk in my opinion and I can't speak to that anymore. Midjourney makes nicer "art", but has soo many restrictions on what words it will take. It's more prudish than the Mormons Here are some words that is has banned: thong, nipples, in the shower, taking off, removing their, etc.
I wasn't even trying to make porn, I was trying to make superhero and fantasy art.
And the image blend is even more laughable. Pretty much anything with visible skin will cause it to give a false positive for nudity. Let's say you take a photo of a bodybuilder in shorts and a drawing of the Hulk in his purple shorts to try and make a new image of the Hulk. You will get a false positive for "possible nudity" and there's really a 50/50 it's either the photo of the bodybuilder or the cartoon of the Hulk that triggered the nudity warning.
It's really only good for making things that have have a highly defined and standard look with many examples it can pull from. I do understand the people that say it's good for concept and prototyping because it can pull from 1,000's of examples and mash something up. However, after awhile you start to notice it has a style or starts reusing poses and faces from before for different prompts. It has a "style" you start to notice. And for as many successful prompts I've made out of refining with variations, there's just as many unsuccessful prompts that I've given up on because it couldn't "mash" up something even remotely decent. And then there's the whole hand, feet, face, and eyes issue.
However, I will say I've found it handy for making "faux" painting with themes or posters that I can use to fill up the background in my renders as props. There's times where I want say "generic sci-fi graffiti with a cartoon cat made with an airbrush" and that does create something kind of generic like that I can them blend into a brick wall texture to make a graffiti wall prop to put in the background in a render. I can't deny it's useful for crap like that. But in that case, I'm still using it to help fill out a render and not the focal point of the art I'm trying to make, which is the rendered image.
Will it probably get better, yeah? Most of the professional artists I know hate it, but they also know that none of this shit ever goes back into the bag once it's been let out. Only thing I think could actually kill this would be large creative companies (movie studios, comic book companies, etc.) getting together to sue the companies that make the technology. But we said the same thing about MP3's 20 years ago too ;-)
|
|
|
|
|
|
Rank: Advanced Member
Joined: 9/15/2012(UTC)
Posts: 591
Thanks: 329 times
Was thanked: 100 time(s) in 88 post(s)
|
I LOVE AiArt - I'd been looking for reliable filters to give illustrated or painted looks for a long time and finally had spent $ to upgrade for new 3D landscape software and the latest Painter but am amazed what AI can do. As 3DCG users we are already used to being insulted as 'non-artists' by people who often have fairly flat, generic styles. The latter are the good ones, there are tons of people with high-school notebook sketch level styles calling themsleves 'Artists' screaming about AI now, people I could have done better than a decade ago.
The cool thing is I can use my 3D skills and sub-professional but not bad art skills to guide it - then my image editing skills to work out the fingers and other kooky artifacts.
What's really sick funny is again the 'artists' that scream its theft then have videos showing how to use Google to look at tons of different art to practice and use for 'inspiration' only a fraction in the public domain. Reading between the lines and seeing some videos I think this Stable Diffusion thing was developed for Adobe. Can't post links here but YouTube > Adobe Ai Art - you'll find plugins including one using Studio Ghlibi style that are 6 months + old and have been active for a year. When the few that knew about AIart thought it was that Deep Dream nightmare fuel of Corgis and Eyes. Also the same search will find a few artists that noticed Adobe had authorized AItraining for years and told how to turn it off... What I think (One Man's Opinion) was that Adobe wanted to find a way to get people to subscribe willingly. Its been a strong rumor assumed to be the truth is that Adobe software has always been easy to pirate and that was a strategy so they'd be industry standard and pros making money and businesses would pay the high fees to avoid lawsuit. But now there' s legit competition (Affinity by Serif...to begin with) and too many "Indie" artists that do gig work vs professionals. Just clamping down on piracy would push them to other platforms and the subscription scheme was hated pushing more demand for alternatives. Having a database for a stable diffusion the "Pros" could get to quickly pull image ideas from and then paint over and pass as their own is a good way to pull $ from high end Pros in exchange for increasing their workflow through the roof and they have to pay Adobe or they cut off that connection so "Pirated" software is meaningless.
Note AI doesn't mean we have literal Artificial Intelligence. Its advanced adaptive software that can 'learn'. Most professional software has such in one form or another and if marketed today would be called "AiARt". I won't do a huge list but my favorite is Corel Painter - which has had "Auto-Paint" since 1998 and clone brushes. You could feed any image into it and have it painted in any brush in its library. Then you could clone brush to add some effects so it looked painted for real vs a cheap filter. Come the 2021 (2020) version they did go full modern AI and some of the filters for auto-paint look much like AIart including fingers going kooky - but based on that one image.
Even with the "For Me but not for Thee" where "Artists" scream its theft for AIs to rove the net and learn from everything but short of Nick Simmons level tracing/tribute they copy EVERYTHING they see save copyright characters and tracing the "Public Domain" has lots of good artists going way back to learn from. Also an 'art style' can not be copyright protected or Mark Bode would be harassed by Disney that his smutty Underground style he inherited from his late father was degrading their "Wizards" acquired IP...no matter who was 'first' but mattering who had the LAWYERS...
|
1 user thanked Sutut for this useful post.
|
|
|
|
Rank: Advanced Member
Joined: 9/15/2012(UTC)
Posts: 591
Thanks: 329 times
Was thanked: 100 time(s) in 88 post(s)
|
Originally Posted by: SnarltheWerewolf
I'm both impressed and disillusioned by it at the same. Stable diffusion has been junk in my opinion and I can't speak to that anymore. Midjourney makes nicer "art", but has soo many restrictions on what words it will take. It's more prudish than the Mormons Here are some words that is has banned: thong, nipples, in the shower, taking off, removing their, etc.
I wasn't even trying to make porn, I was trying to make superhero and fantasy art.
And the image blend is even more laughable. Pretty much anything with visible skin will cause it to give a false positive for nudity. Let's say you take a photo of a bodybuilder in shorts and a drawing of the Hulk in his purple shorts to try and make a new image of the Hulk. You will get a false positive for "possible nudity" and there's really a 50/50 it's either the photo of the bodybuilder or the cartoon of the Hulk that triggered the nudity warning.
It's really only good for making things that have have a highly defined and standard look with many examples it can pull from. I do understand the people that say it's good for concept and prototyping because it can pull from 1,000's of examples and mash something up. However, after awhile you start to notice it has a style or starts reusing poses and faces from before for different prompts. It has a "style" you start to notice. And for as many successful prompts I've made out of refining with variations, there's just as many unsuccessful prompts that I've given up on because it couldn't "mash" up something even remotely decent. And then there's the whole hand, feet, face, and eyes issue.
However, I will say I've found it handy for making "faux" painting with themes or posters that I can use to fill up the background in my renders as props. There's times where I want say "generic sci-fi graffiti with a cartoon cat made with an airbrush" and that does create something kind of generic like that I can them blend into a brick wall texture to make a graffiti wall prop to put in the background in a render. I can't deny it's useful for crap like that. But in that case, I'm still using it to help fill out a render and not the focal point of the art I'm trying to make, which is the rendered image.
Will it probably get better, yeah? Most of the professional artists I know hate it, but they also know that none of this shit ever goes back into the bag once it's been let out. Only thing I think could actually kill this would be large creative companies (movie studios, comic book companies, etc.) getting together to sue the companies that make the technology. But we said the same thing about MP3's 20 years ago too ;-)
With the exception of Anime style you need to download the engines and run them on your own computer -or- rent 'virtual' computers on a server for high pornographic content. There are plenty of paid servers that do let you do spooge even with realistic photos if you do it secret. The best hint is to make them and NOT upload them to the User Galleries - so many do 100 pics before getting something just right they don't have to pretend to care what anyone does to a large point. I won't knock the paid servers their decisions, or get into why.
For your PC the main limiting factor seems to be the graphics card. Need as big as you can afford, my 6Gig is puny vs 24Gig minimum, no care I have 20 cores and 64Gig RAM... Graphics cards are cheaper and I just might rent a virtual PC for now so someone else's server boils water to make a giant spooge picture, at least until I get the hardware. Having said that I've done a lot with mine now that I'm getting these to work and the multi-gig training modules are full of neat stuff. Gigapixel is a good AIbased upscaler that will save you time and money if you want high quality. Yes, it is just better than stretching a blurry pic on a larger canvas, they weren't even a few years back. Works on old porn pics too - might get their version for movies when its on sale again so I can upscale some old X vids too.
Now many AImodules allow input images, some even have GUI where you can doodle stuff.
This is where we 3D people and amateur artists shine - it will often give crazy results but somehow match what you put in to catch the eye...
BTW - if you look you can find Midjourney like modules. They are the best for "Artistic" but way too policed - even private image option is monitored and input images put through too much scrutinty. But again your PC (or rented virtual PC) you can find a similar module if you look hard enough. "Aitrepreneur" via YouTube is one of my favorite sources!
|
|
|
|
|
|
Rank: Advanced Member
Joined: 9/15/2012(UTC)
Posts: 69
Thanks: 39 times
Was thanked: 29 time(s) in 21 post(s)
|
Everything a human can do, robots and Ai can do faster, better, more accurately and cheaper. That is why we build them. And in the end they will replace us. Just as they have already replaced us in many areas.
In the future, there will have to be a different model of society in which everyone is granted an income independent of work, which they can use to make a living and afford all the beautiful things invented by the Ai and built by the machines.
Maybe there will be a small counter-movement that prefers "man-made things", just as there are people today who prefer "hand-made" things. But the vast majority of goods, commodities, art / music will be thought up by Ai and built in a fully automated way. Even political decisions will be made with the help of Ai.
Luckily I am already in the second half of life, because to me this sounds like a distopia instead of a utopia. Billions of unemployed people who see no perspective and no meaning in their lives.
Why should you still learn something, the ai can do it better anyway. why should you still build something, the robots do it faster and more accurately.
You make music? Good for you, the Ai does it too, and as fast as 500 million composers at the same time. No matter what you come up with, the Ai will have already done it.
To me, it sounds like civil unrest, distribution fights and violent riots.
Sorry, I don't want to sound pessimistic, but I work in mechanical and plant engineering in Germany and I see first-hand what the AI-controlled systems can do.
Translated with DeepL 
|
3 users thanked G315t for this useful post.
|
|
|
|
Rank: Newbie
Joined: 5/2/2019(UTC)
Posts: 354
Thanks: 2036 times
Was thanked: 209 time(s) in 179 post(s)
|
AI generated images are just the latest example of technology applied to some human activity. Historically, people have been afraid that technology will make their skills obsolete, and in many cases, it does. The term "Luddite" refers to people who resist/reject technology. Its origin is from a movement of English textile workers in the early 1800s century who destroyed textile machines because they believed the machines would destroy their livelihoods. The world did not end with textile machines, people moved on to other endeavors and textiles became cheaper and more uniform. AI generated art will do something similar.
|
|
|
|
|
|
Rank: Advanced Member
Joined: 9/15/2012(UTC)
Posts: 591
Thanks: 329 times
Was thanked: 100 time(s) in 88 post(s)
|
This - is where it needs to be at: - Universal Basic Income
The "Art" types LAUGHED at weeping truckers, despairing clerks and even went around insulting WRITERS in public...
Souce - my own EYES... A year ish before lockdown. Was in my fav coffeshop/bookstore going over notes for stories I was working on.
"Hey, you a writer?" some kid from the Debt Factory asked. You know, modern college.
"Yes?" I said hoping for some chat, maybe a new buddy or something I could write for $.
"Ahahaah you are a FOOL! AI Writing is going to take your Jeerb!" and he and his buddies danced around saying "Take yer Jerrb!" boasting they were professional artists... Now I could have ripped the ARMS of the chinless wonder if I'd wanted to but while I'm in a "Castle" area that wouldn't be allowed. And hey, all artists, musicians, writers ARE Fools. We could earn so much more $ so many other ways but its burning in us... I looked into AIWriting but still infantile. Chat GPT now is awesome but if only it was as good as AIArt is...
So it's pretty funny - "Don't reality TRUMP fiction!?" that AI is ruining artists before the already ruined (by re-make culture in Hollow-Woke and publishing) Writers were made obsolete. Serious, any advanced thing is $30 a month and for $8 a month you can hire a 3rd world worker. Recent scandal in that lots of chosen for "Diversity" top writers were using ghost writers who sub-sub contracted and the Chinese ones blatantly plagarized. That was ZERO "Ai" involved...
We need to use the mechanics of our Republic and demand Universal Basic Income. cuts to welfare went straight to 1%ers not to the public. Cuts to NASA went to 1%ers bank accounts, not back to the public. Jobs shipped to China and other places were paid with tax breaks and subsidies. And there's still the "War on (some) Drugs" - a lot of programs could be cut to give a huge percentage of the population a free ride modest low end cushion so they'd never have to work 3 jobs to keep barely above starving and have the same place to sleep atnight. IF we had the political will versus fake politics to fight over "Gender Neutral Potato Head"
|
|
|
|
|
|
Rank: Advanced Member
Joined: 12/7/2015(UTC)
Posts: 83
Thanks: 63 times
Was thanked: 47 time(s) in 36 post(s)
|
Originally Posted by: Finister
AI art is... at this moment... based off of a limited number of artists work so the majority of AI art is going to look similar right now.
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.
All galleries could begin to get awash in generic art, drowning out the individual... OR... will this open up an opportunity for the individual to stand out from the generic norm?
What do you all think about AI art?
It only "looks similar" in the same way the "Posette" looks similar to someone new to Poser.
As a longtime 3D user - who frankly struggled to "get what I want" out of applications starting with Lightwave, then moving on to 3DS and Blender -- I _love_ Stable Diffusion. You can "click and make a nekkid girl" . . . and that will look a lot the same and not very interesting.
. . . but if you want something that's got some life, you're going to have to wrestle with all sorts of complexities. I find Stable Diffusion both a lot of fun, and it fits much closer to my creative process. I know that I love old Tom Swift book covers and Helmut Newton photos. With Stable Diffusion, some artfully crafted prompts, and then a lot of editing and iteration -- you can get something that frankly would be hard to do in 3D. Not impossible -- I like volumetric lighting and Mario Bava style cinematic looks, and that can kinda be done in iRay, but slowly and not really iteratively.
So yeah, AI is coming. Its already amazing, and its going to be more amazing yet.
I would expect applications like Daz and Photoshop to adapt, because they're complementary, not really competitive. See, the thing is - its very hard to get a predicatble anatomically correct pose out of SD, particulary NSFW scenes. We've already got an SD inpainting plugin for Photoshop; I'd bet you dollars to doughnuts that Daz will be working full bore to implent AI features in Daz Studio.
. . . and lets' face it, they should do. Because what they've done recently, pumping out ever more seemingly identical and not really compatible figures -- that's not interesting. Genesis 9 is better because? Its not.
So modellers -- don't despair. There's a lot of potential synergy between AI and "traditional 3D"
. . . and for folks earning their keep making erotic models, absolutely there will be a market for customized erotic AI models. At the moment it is _very_ hard to get most of the models to do what you want, erotically speaking. Very simple blowjobs and fucking, basically. Its not at all hard to foresee folks selling models trained on, say Sybian scenarios and so on.
So yeah, we've been using a lot of the same techniques and models since early V3 days -- and everyone gets how ancient the mesh technology in Daz is . . . not how you'd build a model today, if you started from scratch. So there's a heckuva a lot of room for innovation, and room for creators to make money too.
Just my two cents . . .
Edited by user Friday, February 17, 2023 9:25:50 PM(UTC)
| Reason: Not specified
|
2 users thanked topspin for this useful post.
|
|
|
|
Rank: Advanced Member
Joined: 9/15/2012(UTC)
Posts: 50
Thanks: 25 times
Was thanked: 11 time(s) in 7 post(s)
|
I think AI will be able to be taught how to model a 3d model; how to texture it; how to pose it; how to render a story with it; what lights to best use based on the drama of the scene; how to create a cover with titles for the story; how to market the story to the target audience; and AI won't post crazy shit in forums and on media; AI won't feel insecure about what its made; it will take feedback from the hits and comments it gets on its work and make changes based on that for its next work. It will learn what the audience wants and try to always please the audience. AI can only learn from what is presented to it.
It can't be taught to desire. Humans desire to create. AI only creates because it's told to.
|
|
|
|
|
|
Rank: Advanced Member
Joined: 4/21/2016(UTC)
Posts: 318
Thanks: 50 times
Was thanked: 88 time(s) in 63 post(s)
|
Originally Posted by: topspin
...
I would expect applications like Daz and Photoshop to adapt, because they're complementary, not really competitive. See, the thing is - its very hard to get a predicatble anatomically correct pose out of SD, particulary NSFW scenes. We've already got an SD inpainting plugin for Photoshop; I'd bet you dollars to doughnuts that Daz will be working full bore to implent AI features in Daz Studio.
. . . and lets' face it, they should do. Because what they've done recently, pumping out ever more seemingly identical and not really compatible figures -- that's not interesting. Genesis 9 is better because? Its not.
So modellers -- don't despair. There's a lot of potential synergy between AI and "traditional 3D"
...
I'd really like to agree with you there (and I want it to happen), but I just don't see it happening. Daz has been really slow to adopt a lot of things like strand based hair or volumetrics. They only have liquids because of a vendor plug-in. Still no rigid body physics (or any kind of physics), animation timeline isn't great, inverse kinematics is buggy, not even an auto-save feature.
If you look at the market overall, all Daz really has to do is stay slightly ahead of Poser. From their perspective, there's not much point in taking risks and spending R&D money on new features when most Poser users won't switch anyway (they're on Poser because they've been burned by Daz in the past and won't come back).
The ony thing Daz has gone full bore on lately was rushing to launch NFTs while the rest of us were struggling to find GPUs during the mining rush. I haven't spent any money in their store since. In all fairness, that was probably dictated by their new parent company, but the fact remains.
Blender, Photoshop and GIMP on the other hand, have a shit-ton of SD integration. If I could convert my Daz Studio library over to Blender, I'd do it in a heartbeat, but it's a significant learning curve. The main thing keeping people on DS is its ease of use.
The only thing I can ask users to do is look at DS as a free-to-play game. If the company pushes out new content (that costs money) way more often than they update the base application (the free part), then maybe it's time to look elsewhere. I spend way more time in SD than DS nowadays.
sorry for the rant
Slightly offtopic: Can we talk about the non-AI content saturating the AI gallery? Seriously, if your image is mostly 3D based, please provide a comment explaining what AI it involves.
|
2 users thanked Zaavaleta for this useful post.
|
|
|
|
Rank: Advanced Member
Joined: 12/7/2015(UTC)
Posts: 83
Thanks: 63 times
Was thanked: 47 time(s) in 36 post(s)
|
Originally Posted by: Zaavaleta
I'd really like to agree with you there (and I want it to happen), but I just don't see it happening. Daz has been really slow to adopt a lot of things like strand based hair or volumetrics. They only have liquids because of a vendor plug-in. Still no rigid body physics (or any kind of physics), animation timeline isn't great, inverse kinematics is buggy, not even an auto-save feature.
If you look at the market overall, all Daz really has to do is stay slightly ahead of Poser. From their perspective, there's not much point in taking risks and spending R&D money on new features when most Poser users won't switch anyway (they're on Poser because they've been burned by Daz in the past and won't come back).
You may be right about Daz. I have been seriously underwhelmed by them - basically milking the same very old technology. The most important change in Daz for me in the last decade was iRay integration, and that made a huge difference to render quality and speed. Of course, that was really Nvidia's doing, but the integration of iRay into Daz Studio, for no charge, that was a big deal. If you look at the hassle of using the external GPU renderers, like Octane, most people just weren't willing to hassle with it, or pay for it. So that was a good call by Daz to integrate, but obvously Nvidia made it happen
One of the things I've seen in the 3D workflow is that just as I long ago worked to get models out of Poser into Lightwave, 3D Studio and Vue d'Esprit for better renderering, today I see people trying to get those assets into the Unreal engine
The thing of it is -- a lot of what we like about erotica, ain't just a one shot. Its characters interacting in a complicated way. Daz just doesn't do that - even today with a modern desktop, load up four G8 characters into a scene and watch the whole thing come to a crawl. So the immediate threat to Daz probably comes from Unreal . . . I had been exploring the capabilities there when Stable Diffusion showed up, but whatever the case, hauling around a 100k polygon model to make a picture doesn't make a lot of sense.
Originally Posted by: Zaavaleta
Slightly offtopic: Can we talk about the non-AI content saturating the AI gallery? Seriously, if your image is mostly 3D based, please provide a comment explaining what AI it involves.
Not sure that was addressed to me. My content isn't mostly 3D based, anymore. What I'm doing now is Stable Diffusion . . . I do use Daz to render simple silhouettes to get pose. Its very tricky getting hands posed accurately in SD, and sometimes when I just can get it to behave, I'll pose the hand the way I want it, then composite into Photoshop, then push that output back into Stable Diffusion - using an image prompt as well as text.
There is a brand new plugin to Stable Diffusion that allows for more accurate posing inside SD, ControlNet.
AI applications can presently generate meshes, which can then get rendered in traditional 3D programs . . . at the moment they're not very good, but just watching the speed of all of this, I can imagine that this may evolve very quickly.
I should add that its a pity that this got the monicker "AI" - its not "artificial intelligence", its basically indexing in to a "latent space" of possible images from training on existing ones, with a catalog of labels. Train on other images, with other labels, and the capabilities are different -- but, for example, a SFW trained model won't ever get you much in the way of more explicit NSFW images, it just doesn't have the data to support that. Edited by user Saturday, February 18, 2023 1:19:59 PM(UTC)
| Reason: Not specified
|
|
|
|
|
|
Rank: Advanced Member
Joined: 4/21/2016(UTC)
Posts: 318
Thanks: 50 times
Was thanked: 88 time(s) in 63 post(s)
|
Originally Posted by: topspin
Originally Posted by: Zaavaleta
Slightly offtopic: Can we talk about the non-AI content saturating the AI gallery? Seriously, if your image is mostly 3D based, please provide a comment explaining what AI it involves.
Not sure that was addressed to me. My content isn't mostly 3D based, anymore. What I'm doing now is Stable Diffusion . . . I do use Daz to render simple silhouettes to get pose. Its very tricky getting hands posed accurately in SD, and sometimes when I just can get it to behave, I'll pose the hand the way I want it, then composite into Photoshop, then push that output back into Stable Diffusion - using an image prompt as well as text.
There is a brand new plugin to Stable Diffusion that allows for more accurate posing inside SD, ControlNet.
AI applications can presently generate meshes, which can then get rendered in traditional 3D programs . . . at the moment they're not very good, but just watching the speed of all of this, I can imagine that this may evolve very quickly.
I should add that its a pity that this got the monicker "AI" - its not "artificial intelligence", its basically indexing in to a "latent space" of possible images from training on existing ones, with a catalog of labels. Train on other images, with other labels, and the capabilities are different -- but, for example, a SFW trained model won't ever get you much in the way of more explicit NSFW images, it just doesn't have the data to support that.
No, not addressed at you at all. Sorry, I'm not clear when I write sometimes. Your images are brilliant. Sorry I haven't left a comment yet.
I was saying there are quite a few images in the gallery that appear to be just regular 3D images. It seems like people are just checking both boxes [3D] and [AI] to get as many views as possible. A lot of images in the AI only gallery are also in the 3D gallery. I kind of wish the submission screen made users choose one or the other when they submit.
Models are becoming a source of pain for me. I can only do so much with the SD 1.5 and 2.1 models. When I try to train with Textual Inversion on any other models, the results come out hideous. And I can't get Dreambooth working for some reason (Pytorch dependency issues).
|
1 user thanked Zaavaleta for this useful post.
|
|
|
|
Rank: Advanced Member
Joined: 12/7/2015(UTC)
Posts: 83
Thanks: 63 times
Was thanked: 47 time(s) in 36 post(s)
|
Originally Posted by: Zaavaleta
Models are becoming a source of pain for me. I can only do so much with the SD 1.5 and 2.1 models. When I try to train with Textual Inversion on any other models, the results come out hideous. And I can't get Dreambooth working for some reason (Pytorch dependency issues).
The best model, in my limited experience, is "Deliberate 1.1" -- with an NSFW tag in the prompt, you get very different things to the standard SD model. SD 1.5 will do NSFW, grudgingly. SD 2 seems to have nerfed NSFW. Many of the folks doing hardcore are using the "HassansBlend" model . . . which I don't particularly like, since I'm not usually going for just pornography.
I also do something which I gather other people don't do much -- chain the models together. So I'll render out some images in an Illustrator model, then take those sketches and go into somethign more realistic like "Deliberate".
I am literally learning something new every day. Its an exciting time, and while its a new set of skills, when it works its pretty cool . ..
I highly recommend folks who are interested try out the Leonardo [dot] ai cloud server. Its still genuinely free and in Beta, but won't be free for long. Its much less hassle than trying to get these somewhat wonky apps installed correctly on a local machines
|
|
|
|
|
|
Rank: Advanced Member
Joined: 9/23/2020(UTC)
Posts: 338
Thanks: 75 times
Was thanked: 186 time(s) in 170 post(s)
|
To stay within the framework of Renderotica: I think there will be - after the hard core barrier has fallen - 3 groups:
The group "Just type and it's good".
From this group we can expect pictures like "Orc with big dick bangs elf with even bigger boobs in the forest". Yeah!
or "Old clothed dude accompanied by beautiful half naked 21 year old girl". OK, these are the excesses of patriarchy.
The group "Simple typing, but with intention and soul".
These folks will entertain us with pictures like "Orc with big red dick and a golden crown bangs elf with big boobs and wings in the forest while
the moon shines and a deer roars in the background". Yeah! or "Old clothed dude accompanied by beautiful half naked 21 year old girl". Face palm ;-)
The last group
will explore all AI possibilities with heart, brain, soul and mind and also use post-processing to create wonderful, novel images. Most of the artworks of "topspin" are a good example of this.
Am I - as a graphic designer and 3D CC - afraid of AI? Nope!
Rather, I'm looking forward to all the works that are yet to come ... with the exception of this boring "Old dude" stuff.
|
1 user thanked Pushee-Ri for this useful post.
|
|
|
|
Rank: Advanced Member
Joined: 12/7/2015(UTC)
Posts: 83
Thanks: 63 times
Was thanked: 47 time(s) in 36 post(s)
|
Originally Posted by: rollemops T
The last group
will explore all AI possibilities with heart, brain, soul and mind and also use post-processing to create wonderful, novel images. Most of the artworks of "topspin" are a good example of this.
Am I - as a graphic designer and 3D CC - afraid of AI? Nope!
Rather, I'm looking forward to all the works that are yet to come ... with the exception of this boring "Old dude" stuff.
Its always been the case that the tech doesn't make the pictures - Stable Diffusion with no inspiration gets you a very generic blah girl, looking like Hegre or Metart.
Getting something interesting - requires some inspiration. Over the years there have been amateurs who didn't have great art skills, but were hugely creative, thinking of people like the BDSM artist "Augustine". Very crude drawings, but a compelling story.
My two cents worth is that the AI is very responsive to a knowlege of art and photography. If you say "Bauhaus bondage scene photographed by Richard Avedon" - you get something. Now, that somethign may be shite, but if you iterate enough, have enough of an idea how styles work . . . its a lot of fun, for me at least. It would be tough in a production workflow, at least with the tools that exist now - they great, but unpredictable.
Stay tune, 'cuz it changes every day.
|
|
|
|
|
|
Rank: Advanced Member
Joined: 9/15/2012(UTC)
Posts: 69
Thanks: 39 times
Was thanked: 29 time(s) in 21 post(s)
|
Originally Posted by: rollemops
the moon shines and a deer roars in the background". Yeah! or "Old clothed dude accompanied by beautiful half naked 21 year old girl". Face palm ;-)
Whats wrong with "old clothed dude accompanied by half naked 21 year old girl?" I also have an image like that in my Gallery. :-D
It just shows the reality of that business. 
|
|
|
|
|
|
Forum Jump
You cannot post new topics in this forum.
You cannot reply to topics in this forum.
You cannot delete your posts in this forum.
You cannot edit your posts in this forum.
You cannot create polls in this forum.
You cannot vote in polls in this forum.
|