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
© 2024 Daz Productions Inc. All Rights Reserved.
Comments
Regarding merging-- The re-mixing thread is all about creativity and using Ai as a tool to help that process.
@Federmann, wow, you remember that render? Thanks!
For me, every tool has its use. For someone that, for example, wants to make a comic with a 3D look, Daz is a much better tool than SD, since you can get exactly what you want, have perfect continuity, etc. On the other hand, for someone (like me) that wants a real-life look, Daz is sometimes frustrating, especially when the renders involve people (skin, etc), and SD is way more rewarding, let alone the render time. But, as everything in life, a product that could integrate the best of both worlds would be fantastic...
@PixelSploiting also have a point: the spamming of IA images is everywhere. 3D images forums, "erotic" forums, stock images sites, etc, etc. It's a good thing that in the USA they decided that IA images cannot be copyrighted, except when there are real, significant changes after the generation (that turn it into a personal work). If they hadn't decided this way, someone would start generating infinite images in infinite styles and copyright them all, only to sue the world after that. It would be a new type of "patent troll", but for images.
Something that I don't think has been mention previously on this thread is the cost. Like DS a local install of SD is free of charge, but more significant is that nearly all of the content is also free, all of it free at Civitai, with a very small percent sold through Patreon, Koffee, etc. This makes Stable Diffusion cost-free for users, at least for the time being.
Per se... yes. But you might have to consider the copyright question, if you use a training data set, that consist of images, that may not be licensed for the purpose ~ minus some uncertainties about the regulation concerning that, multiplied by countries and years to come - though i think, currently it looks like training data is going to be regulated "with copyright applying" in biggest places. Feel free to correct me, if my data on this is outdated.
So from there, you either would need a pretty large curated data set that's safe to use, or you'll select carefully yourself, and/or use data sets that certainly are free to use, e.g. if a museum consents with putting it up, like for paintings and art that isn't copyrighted anymore (careful with assumptions, though, because a photograph for instance, could have some kind of protection, so here rule no.1 really is to know what the training data for a used model is. (Yes, for your own fun, you can use virtually anything at home, plus minus some place's laws, but then publishing it is another thing.).
So here i am not sure if cost vs. result is the thing. If you can achieve something specific with training your own models, that's pretty cool. Not sure if there has been any motion or clarification on using daz renders for training, assuming you're using it to enhance/modify your own daz renders, only employing products you own. Last time i checked, i didn't understand the new-ish EULA.
For another angle, there will be commercial data sets with commercial engines, as i assume there are a few already, which appear to be pretty safe to use, though in the early days it still looked like some players still tried to get to use whatever they had on their platforms without asking for extra consent, then special methods of asking for consent... different topic. From a theoretic angle, there will be larger data sets, free for ai training at some point, or maybe they'll just be legal in the process of a certain (commercial or not) player using them. It may take a while for a huge one, but once there, the costs will be determined by the hardware needs for running downloadable models, plus maybe a fee to use a model, due to the cost of training in the one case, and in the other case by whatever they had to do to achieve it, for closed commercial cloud models. I am not sure how cheap that will be, in the end. Cloud pricing typically has one direction: up :). Correct me if i'm wrong for the present days...
Concerning the "ai" and 3D, like with DAZ Studio, i'd have to add, that ai isn't just ai.
We have the generative models now, afresh, but also some kinds of more well-seasoned approaches, each of which could play a role in enhancing Software like DAZ Studio. Some applications are already existent, but probably not trivial to train, some can clearly be envisioned, and there very well may be a lot to discover, still. Generative models can also be used to enhance other different models, e.g. during training phase, so there is a vast range of things that could happen.
The impact on DAZ really will have to do with what they're making money with, and if they can keep maintaining stuff (and want to). My assumption is, that they'll have to add some tools that will drastically simplify some tasks, in order to keep up. Don't ask me for a time frame.
In theory, 3D art with precision likely will not become outdated, until something can actually replace all it's core features pretty much 1:1. That simply because "ai" is also used to anhance workflows in 3D in various forms, up to placing and posing and animating (and lighting, and ...) becoming pretty easy-ish, in terms of barriers. Barriers will fall, like i can describe what i want, or at least where to put something, without bungling one of three axes all the time for minutes. But increased effiiciency most of the time means that some things will sell less. Sometimes some things sell more due to it, but sometimes just the tools sell more. It needn't be bad, but it's hard to tell how transitions will look like, and to where exactly they'll lead. It's possible to make mistakes, and to find a working way for things, of course, in a phase of uncertainty and so on.
I wish they would just train a Stable Diffusion Checkpoint model entirely on Wikipedia CC0 images and be done with it
donations to Wikipedia for downloading it would be nice too
(and a paid online web image generation option benefiting Wikipedia for those without a powerful GPU)
since they are captioned the results should be very coherent too
this would solve the copyright and ethical issues while also providing a great resource for educational and recreational use
I did a render of a character with clothes, put it on an AI created background then flattened, did some post then put the final image through AI and got some interesting results. It's possible to do both. I also use AI selfie apps on completed renders. I think it's a great tool but not going to replace 3D yet... There is still a lack of total control using AI but it could improve in the future. DAZ/Tafi is doing some kind of AI now too, I saw on Twitter but haven't really looked into it.
I am positive that there will be pretty large and free to use sets, eventually. It's just not possible to do it quick, if you don't want legal risks. Perhaps there will be some kind of image license (add-on), that allows for training non-commercial systems in the future, somewhere. But for a license that explicitly (or effectively legally) allows use for ai training to be widely adopted, it may still need quite some time. Regulation certainly isn't done with the topic, and who knows where we are going in the end.
(Concerning the solving: some ethical question could actually remain for the whole thing, even if you are relieved from all the questions around using it. That's more like a societal thing, e.g. if it happens to lead to destruction of something we need. Such could happen, due to the way most people use the capabilities, in the context of how copyright law and how automatic filtering works, then. Further we might see commercial players trying to tip something over the edge, for their own short-term benefit. So i still see some future haywire potential for the general case, even if it turns "all democratic" with less hardware requirements and free and safe to use training data sets.)
well there are always going to be questionabe uses and illegal content produced like people do already with the pen and paper, the typewriter, the camera
and 3D content, DAZ implimenting making bits disappear as characters are aged hasn't stopped those who render bad things.
I render horrible images. Regarding AI art, what if it never leaves the website it was created on? Would everyone still be upset? I play around with it but have never posted any of the AI's work on any website; I like it when you put Haiku in the prompt. There's no telling what you come up with. So, what could it harm if someone keeps them separate from the Purist View of 3D art? My purist view on 3D art is that there is no postwork; the only postwork on my images is my name. Again, I play around; I don't take it seriously as a hobby. I don't try to pass it off as my art. It belongs to the AI.
I also render horrible stuff, over and over again:) Hm..where were we.. oh yes, AI... is that short for anti-ism? If so, let's start a Stop Anti-Ism-Ism before it stops You - movement and go all in!
Isn't this another trade off between ease of "creation" v. control. Okay, so It is easier to get AI to generate an image than to paint one yourself, but is the resulting image what you had in your mind's eye..
You see all those slick looking images, but are they anywhere close what the artist was trying to convey? I doubt it. Most of those images seem like they are happy accidents that occured while struggling to get it to do what they wanted. Reminds me of the good old days of playing with .Mandelbrot generators.
Then there is the sameness of all those images. The images seem all shiny at first, but after people get used to seeing AI produced art, I suspect they start to pick up on the cues that scream "AI" and subsequently will become bored with them.
I didn't mean that on the personal level, except maybe for "CEO-level". Rather consider potentielly unwanted or dangerous societal or just large scale effects.
Concerning the CEO, one not so modern fear is, that some players abuse euphoria of people to make them a weapon, and just toss the whole thing into the abyss (potentially unintended effect). There are various dystopian potentials in the situation, e.g. a world where sellable and protected artwork only exists by the grace of a data-pants-down registration approach, while everything else is free for the hunt. (Classics of businesss architecture.)
Otherwise: Pen and paper are still not (fully) out :).
AI won't be killing anything any time soon, especially considering what people currently call "AI"; basically all it boils down to is an enhanced database / search engine which can't really piece anything together on its own. The main difference is that AI has now become a sales model and thus... companies start to overhype it. And honestly... no offense intended but discussions like these amuse me because it becomes apparent that not that many bother to look deeper into the subject at hand.
See, the reason why I call it an amusing & overhyped sales model is because there's hardly anything new... the audio industry has already been there.
3D is a hobby of mine, but digital audio has become a passion of mine over the years. Yah, AI has long made its entry within the industry years ago, of course back then it wasn't touted as AI but merely called for what it was: smart automation. I have several instruments which span a large variety of audio: percussion, bass, guitars... and some of them take the automation part extremely far: even up to a point where the instrument can pick up tempo changes and anticipate for those all on its own; based on the generated audio vs. tempo settings.
And that't not even addressing the extensive options within audio processing; audio effects which can enhance material solely based on the material itself: automatically detecting basses that can be beefed, picking up high notes that can be settled, stuff like that.
Sure, this had its effect on the audio industry but it never killed it. Why would it?
Same will apply here. Heck... IMO one could even argue that Mondriaans rather simplistic designs could be considered to be "generated" yet people still enjoyed that as well.
Just because there's a price tag on something, it doesn't mean anyone is going to pay for it.
Everyone can afford a pencil. How many people can use the pencil to write like Shakespeare.
Everone can instantly produce a photograph, 99.999999999% of them nobody wants to see.
Everyone can almost instantly produce a professional looking website or blog, that doesn't mean you're going to get any traffic.
Using tone-correcting software like AutoTune, your singing voice can be made to sound perfect. How come you're not making $100,000,000 a year as a music superstar?
Video killed the radio star - and streaming killed the video star.
When AI can cook me dinner, serve it, then clear the table, and wash the dishes and put them in the cupboard - I'll be impressed.
AI is already cooking: Humanity is dinner.
@ShelLuser There is something new with "generative ai", which appears to work out in an interesting way, from a certain model size on, given a few other improvements (both text/image). Certainly not due to some 5% increase in computing power. This seems to be a difference to most of the examples you made, as the models from years ago are likely of a different kind/size. though generative ai in general, isn't that new either. It has been researched for a while, and there have been applications running in various fields (likely pharma, google internally, others). I'd call the hype a hype as well, but the field shouldn't be underestimated. The "killing" might happen with "killer applications" ;), entirely for commercial reasons, i.e. price tags for results, e.g. if politics fail to regulate (short version, either destructive potential or significant improvements still would have to unfold somehow). This won't kill the full precision pipelines in general, open source won't just so die, movie people will have the best gear anyway, and the commercial 3D players will adapt somehow, possibly to varying degree of success.
Just throwing this in, because i think "the ai isn't the ai".
Slight objection :). AutoTune isn't necessarily that cool, even if you "correct" very good singers. There's very instructive examples of very good singers, produced with auto tune, killing off a lot of potential. So i would object in that sense, that "sound perfect" isn't the case. AutoTune itself may "pitch" it perfect, but we typically also don't prefer clean sine waves or chainsaws all the time, so unnaturally straight tuning will add to uncanny at least.
I tried Nightcafe Studio (because I've been collecting 5 free credits for months now) and AI is definately going to be a threat to DAZ 3D, Poser, and surprise, Blender and those type tools. Not that they won't be used by long time professionals anymore, but that they'll siphon off new adopters of those tools. It is still extremely difficult to guide the AI to make exactly the art storyboard you want, I can't do it sufficiently to satisfy myself at least, and that will have to be extremely easy for people to actually use it for serial production of a story usual AI long term and not just some easy way to 'press button - make pseudo-random art' on a topic. I like it enough that I'm going to sign up for their Pro level at the cheapest basic rate (i.e. $15 a quarter which will allow me to use my own datasets to try and guide/customize AI storyboards better, eg using my own DAZ renders would be possibly).
And then when I look at the entire gallery of others - they are ALL so very, very good and virtually effortless, even if they all have a same-same quality to them because you know they are AI in the same way you know a 3D CGI render is just that.Even with that, there is still no denying they are good, very good.
AutoTune has a "humanize" adjustment to make the perfect tone slightly imperfect - so it sounds human.
Therein is the problem. If I want to see beautiful art, I can go to a museum. If I want to see great photos, I can go to any number of places that stockpile photos of anything and everything. If my main goal in creating art is to create a random image and stand back and say, "Self, that's an awesome picture you just clicked a button to make," I can do that with AI. And there's nothing really wrong with that. I can totally see how someone who doesn't have a specific image in mind (or the need to make a series of consistent images) may well go AI rather than Daz, Poser, etc. You can get cool, random, one-off images about a general subject with no investment of time and money and that's a way more sensible option if that's really the goal.
If you are the process-oriented type who wants something just-so, but doesn't start off with a specific idea of what that something is, AI could still fit that need. We hear about people spending hours refining prompts and previous generations of images in an attempt to do that.
But if you want specifics, AI is still not there, as far as I have seen. Will it be? Two years from now? Next year? Next month? Who knows. But for now, granular control over every aspect of a scene and the objects in it is what Daz (Poser, etc) offers and AI can't deliver that. It's still two different tasks, and two different tools.
Yes, but the singer is still better, if they are good. Character of the voice is made up by details, not just humanization, then there is all the slopes and slides, and details of expression. I just wanted to (slightly) object the perfection. Sometimes producers overdo stuff. Many plugins are better used at an optimal dose, whatever that may be for a specific case.
If ai could "win" via perfection, we'd only have the best of singers in the charts by now, shouldn't we? That obviously hasn't been the case (~ever), despite all the professionalism added during recent decades. AI could be used to mimic some version of "successful", including all the mix and choreography. Luckily half the people tend to hate the other music virtually always, so we're likely good.
And yes, the barrier will be muddled anyway, and they'll be able to mimic styles and all. However it'll often come at a cost, because you get the details "right" in style X rather than fits the specific song. It'll always tend to stay somewhat statistical plus maybe smoothed out a little. So if you go into the details anyway, you'll also enter the hell of details anyway. With ai we might get the sculpting tools. I'd be happy with something a little more serious than vocaloid 5, vamped up to being able to sculpt a (preset) voice a bit, rather in terms of emotion, resonance area emphasis, actual singing parameters, with a more intuitive and spoken interface, you name it. Especially if i'm enabled to preset "a way of singing for a character", i.e. rules for when roughtly to do what. Rules needn't be "programming" it could be me just emphasizing the aspect with my voice, disregarding all other parameters like pitch, and the system will use the voice appropriately (not pitching/bending my voice). If ai leads there, it'll result in a tool, rather than a (well done but still) smoke screen. Ok that went slightly off lane, with constructing another voice, rather than enhancing yours. We might, though, get to a point, where you could set up such a preset, to mimic your own voice, and even incrementally add detail, whenever needed. That'd be fun too.
Let me put this differently then... IRay is a photo-realistic render engine. In fact, my girlfriend (who is plain out passionate about 3D; that's where I got my hobby vibes from)... anyway: my gf can somewhat easily set up a render, fully MDL, after which you may honestly wonder if it's real or not. Computers have come that close already quite easily.
Question: Did this digital photo-realism kill the need for human models to pose for images?
Answer: Nope!
Why would this fare any differently?
The lack of precise control is what made AI art inferior at moment, but I'm amazed at the speed it improves, a few month ago most AI models such as midjourney and leonardo struggle to generate human hands and eyes. now this problem seem to be over came with good promts. These days I use these AI tools to improve my 3d renders, such as denoise, and change color tone, they work wonder, and much faster than photoshop
Will be interesting to see how an agreement with Hollywood writers will affect AI usage at studios.
It said writers got most of their claims, but studios could use their already own manuscripts to train AI.
...and maybe train some bots to go to the movies to actually watch the stuff
For the posing: Digital Photorealism has been and probably for a part still is, tedious to do. A professional photographer might work in very efficient ways, and the interaction with a model may give you that extra bit of situational potential, which keeps your ability up, to produce something new in an efficient way. Further, many of the (post) tools in 3D similarly exist for photography. If Image generators make the prices drop, paid models might be gone the other day, at least for the realms, where there is pressure on the pricing. But perhaps better posing tools for 3D-Software would make this happen already. It'll have some delay always, because some people stay with their workflows for a while and customers might be hesitant to switch or experience outright failures from early adopters of image generation.
On the other hand, did you actually check the numbers? I am very confident that photography has made the number of jobs for "posers" dramatically increase, because literally no one can paint well, while everybody can hold some camera - this hasn't changed over centuries, except for cameras not having been available until their invention. Hardly anybody would pay a painter to paint them anymore, though.
Then you have the aspect of stories. Influencers and people who provide some sort of content, stories. These are more these days, but they're their own models, in a way.
But for the actual argument: I don't think 3D wil be gone (ever), because much will result in tools to be better/faster with 3D, and precise control of results will always count somewhere, which even with "one ai" would literally still have to mean "3D".
I'd rather assume some bankruptcies and buy-outs medium to long term, due to economic pressure resulting from new tools with/around image generation. Especially if regulation fails, some (bigger) players could try to empty some bath tubs, leveraging their access to data and number of users, having the data centers and models in place. Much of that would go along with manipulation and excluding others from markets, e.g. using or creating a hype and acting fast and deadly. Interfaces similar to what Adoby Firefly looks like, with lots of control and step-by-step creation, could lead to relative powerful and fast tools, though many steps in there are generic in terms of just being tools within a process. Tools that others can build too, in general. Such could also be applied to 3D software, becoming integral parts of such toolchains, so there is no long-term issue with tools, however integration into complex existing systems isn't trivial, and i see some potential for a phase of (up to much) less revenue for those players, who somehow fail to adapt. Another factor could be that the users just cling to what they like to work with, because there is no reason to switch - this would only change with economic pressure, forcing you to be cheaper. Thus if their tool fails to become that bit more efficient as well, they'll be forced to go to another software or drop creating images in that way alltogether. Trying to adapt has risks too, so i expect difficulties for some players, who adapt wrongly or in an unlucky way (it can't all be foreseen).
The abundancy of generated art might have consequences too, leading to various directions (just dropping prices, regulation going wrong or just missing, dystopy of control, no copyright anymore, new courts of nobility, ...)
I've been trying to make a shorter post here...
Perhaps it depends on two big points:
- Foreseeing consequences and regulating swiftly, meaning: it needs to be regulated, and you can't wait another decade. (Also and probably even more so for text.)
- Toolmakers adapting to the situation, which means to understand what actually is the situation, improving their products as necessary.
- Toolmakers adapting to the situation, which means to understand what actually is the situation, improving their products as necessary.
I hope to someday have you reveal that most of your posts were made using ChatGPT or something.....That would be wonderful.
Using AI to debate about AI....
so, because ANYTHING that looks good is now accused of being made by AI, it's now policy to constantly SHOW the Daz Studio interface.
I never thought I'd have to cite Daz Studio as a tool for Artistic Authenticity. lol
One of the issues I have with Ai 'art' is that it is a dead end for developing any skills for the end user. I like to learn and progress, and I definitely value things I struggle to make more than things which were very easy to make. Beyond learning a few ways of writing prompts I don't see any skill or opportunity for improvement with using Ai. The engine might improve, but that's not something the end user learns anything from.
Possibly it depends on the reason people produce art, if it's for 'likes' and comments like 'ooh that looks great', then yes I can see the value in making Ai generated images.
One question about (commercial) AI is: "Is re-iterating the good enough good enough?" - Imagine that scenario for reference, meaning the entirely encyclopedic reproduction, in terms of some synthesis, some fusion, lots of improvements and integration with other tools for an interface. So we boil our artwork-side down to "a million different things", and don't really have an issue with that, because we don't even remember the last 5000 ones. In a way, asset stores appear to run similar strategies at times, however there still is technical improvements over time. (In economy 1 million things won't suffice "ever", because someone will find a blind spot, and shove everything into the abyss the next day. Fancy wall vs. Firewall thing.)
@Griffin Avid Wonderful! Pick your favorite continuation:
1. I'll refrain from answering that one right now.
2. All this has happened before, and all this will happen again: https://en.wikipedia.org/wiki/Mechanical_Turk
3. I've tried so hard, but in the end, the post just had to collapse under the weight of the very letters letters, it had been made up with.
@background For the masses of users, it might even increase skill, if they busy about construction details for the first time in their life, or if they happen to value actually well made images, due to using ai. The argument that for the majority of creators using ai, also for professionals, we'll probably see a decline in actually useful skills, may well be a valid one. Though it might happen, due to people joining in, who don't have the skills in the first place, statistically reducing skill at first. Really depends on how ai plays out, e.g. commercial cloud systems mean no control over anything anymore, and no certainty about future compatibility. Instead open-source open-data approaches turning out to be fully usable at modest hardware requirements, or open source foundationes creating working counterparts, could mean a different direction. Some report fear of ai, from some of the big tech people because it threatens to become open source effectively, technically.
So for both topics, "in the end" there is a lot of uncertainty, and while not being a linear thing, it'll certainly have an impact with a pretty large "up to..." part, for an educated guess, rather than fandom. Staying informed and considering (rather specialized) professional's advice continually, should make sense for the "tool makers", because things can move very fast with this kind of technology. Don't panic, though.