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
@generalgameplaying I think Eliza did it better tbh.
Generated AI:
I type in a text describing what I want, and I can choose from some results.
3D:
I create/ modify a 3d mesh and map it until I am happy with it.
Feels like the difference
a) between online shopping and going into a store, where I can actually touch things, before I buy them.
b) of download songs and having a vinil album of your favorite band,.
c) of downloading a movie, and having a night out with friends in a movie theatre.
Not sure if you intend to focus on the haptic difference or the piracy aspect. Concerning what we endorse or cause with participation:
A technical difference is that generative ai needs images as input for training data, so it feeds on the past toolchains. It can't be built without that input, similar to daz studio not doing much, if you don't have any assets. But i think we shouldn't do with too high levels of abstraction, because generative ai outputs something close to an end product, under a coarse view, takes "no time" for it, feeds on artist's work, which has been created in more classic ways, while at the same time posing a direct competition to them, and can not feed on it's own output, conceptually.
The abilitiy to mimic and modify, the encyclopedic nature of current systems, scaled to millions of users at a high output frequency, means a different beast.
(For those hoping for a database with safe and sound copyrighted art, i may suggest reading the news about "state of the art"/nowadays content id system on youtube, which is being abused in an attempt to silence certain creators, possibly featuring further ramifications concerning control over other people's content. It'll be fun for sure, especially, if you can't even distinguish that well, what has been created by ai. Trying to mend that... i've written about some odds and ends in other threads.)
Got a survey from Daz wanting to know why I had stopped buying content and one of the possible choices on the form was AI tools as the reason so maybe at Daz HQ the light bulbs are turning on.
Me too;) Will check it out later. Over here I can say that me not buying content has nothing to do with anything AI-related...
Hmm, maybe I should stop buying altogether so I can get the same survey so I can tell them I want nothing to do with AI when it comes to rendering and art since the negatives far outway the positives in my view. For the record, I am not anti AI, I can see the benefits of AI generated NPCs in games and advancement of medical options, but not when it comes to training it to create art from photos and other sources it shouldn't have access to.
the last line of your post is actually the only thing I see as an issue for AI
not a big enough issue for me not to play with it myself on my own computer and free sites
but definitely why I am totally against the commercial use of AI art
I won't get an email because I buy way too much DAZ content too
I even render some
AI won't replace 3D but a lot of PA's are going to wonder why their sales plummeted because they couldn't step up their game. Image below is Floyd with an AI generated background.
As pointed out earlier, this thread is for discussion only. If you want to post images there is a thread for that
https://www.daz3d.com/forums/discussion/591121/remixing-your-art-with-ai/p1
Even if PAs created the most realistic assets, some users are still going to opt for an option where they do less work and get great results due to it pulling from parts of everything on the internet including real photos used to enhance an image, so no matter what a PA does, it will still be a loss for them
...for myself it's G9.. Can't justify dumping more money on my meagre income on a new figure to get it to the point where it is as useful as what i already have. I still buy stuff but it's limited to props, sets and utilities
...bingo.
As I've mentioned before it's essentially the "Make Art Button we've joked about here for years..
I didn't get the survey at all and my content purchasing has greatly decreased as well in the last year & a half. I can only guess you 2 are purchasing even less than I am currently after purchasing much more that I even did prior.
I actually this month subscribed to NightCafe.Studio quarterly at their $15 entry level pro. I have used presets and gotten one or two AI renders that I might actually print out and promising "cute" toony stuff that is unusable because of AI shortcomings that I might could maybe erase the mistakes if I learned the proper text prompts and/or moybe needed evolutions to correct the mistakes. I have no ideal yet how to control it enough to get render I want out of it but I will watch some tutorials on how to use it to at least get closer to that goal as everyone else's AI render posts are clearly 'under more control' (i.e. better than mine) by the user than mine are. To do that I had to subscribe to the pro quarterly (I buy basically 100 credits for $15 every 3 months).
..that won't be very hard this month
...the one major issue it won't be able to handle is originality, be it a character, an item or a unique setting one has in mind.
I could get it to generate a young looking ginger haired girl, but it won't be Leela the way I envisioned her. Or say a cool futuristic car but it won't be an Æon (a company in my story world) Zeus limousine.
That is where AI generated art fails
Actually, end users can train the AI models by creating Textual Inversion Embeddings, LORAs, or Dreambooths. These are like "morphs" that edit models in Daz Studio, providing new information that modifies the base AI models. You could train one of these on a character you made in Daz Studio, by using reference photos that are similar to the concept you want to train, or by prompting an AI tool to generate the person based on the description. Now these won't generate the concept immediately, so after training the concept, use it to generate more images and pick out the ones that closest match your idea. Feed these back into the training process and create a new training, use it to generate more images, and slowly you can converge the training to the concept you want to train it on.
...sounds easier (and more rewarding) to just develop a charater form scratch and save it as a character preset.
Bravo! Well said.
I much prefer 3D to AI as a tool for making consistent illustrations. With AI you get different results every time, even if you have a picture as the base. I use DS as my primary illustration tool, because it lets me transform my vague ideas into visuals through trial and error, in a way no other medium has been able to. What I can see myself doing, when I get slightly more of a hang of the free AI programs I've been playing around with, is to use them to generate character ideas I can dial a 3D model from. My friend and literary partner-in-crime generated some pretty good fake photos of the main characters in a novel we're working on. It's going to be so much easier making Dazzies of them with those for reference.
This!!
In my opinion it's a combination of the following
- it's easy too create nude women with AI without having to learn a 3d program and buy stuff.
- Generation 9 is quite resource hungry and didn't bring major improvements.
- there seems to be a change in PAs.
- many new products do not have the quality of previous products.
Example: I reported an issue with Amelia 9's make-up. First reply from support was that they could reproduce the issue and it has been added to bug tracker. Next day, the issue is set to "works as designed".
I see AI art versus 3D as like comparing fast food with gourmet food
for many the former will do, is quick and convenient but it probably is not great for your health, leaves you hungry etc
Wow, an AI which would assist iray by saving render time (and electricity), i would like that, and it would be environmentally friendly! That might even help 3d art.
It depends. If you are trying to replicate a real person and you have a good set of source images, it is way easier than trying to create the skin textures, create the morphs, and then realize you can't get a good likeness because HD morph creation is PA only. Given the popularity of image to Daz character tools (Face Transfer, Headshop, Facegen, etc.) there is a high demand for this workflow. Also, for skin tones where there is little Daz representation, you are limited by what PAs have made available and how well you can edit them to match the right look.
It's not just people concepts you can do training on. For people who don't have great modeling skills it can be a great way to create clothing, props, hair, etc. I was able to take screenshots of an outfit from a cartoon, train a LORA, and then apply the LORA to a photorealistic model with about 45 minutes of prep work and 30 minutes of GPU training time. Stable Diffusion was able to take the cartoon outfit and transform it into a real, photoreal outfit. Creating the same outfit in Marvelous Designer, rigging it in Daz, UVing, texturing, etc. would have taken dozens of hours.
If you are going to use Ai to create a likeness of a real person, then you might as well use Photoshop ( or some other image editor ), to put the persons head on someone elses body, either way it's a fake.
Once you have a character LORA or Textual Inversion Embedding you can basically use it like a Daz character. Posing, setting clothes, accessories, hair, etc. This is way faster than creating a fake in Photoshop. A photoreal picture can be made literally in seconds. You can even take these trained characters and use style transfer so you could generate an image of that character as if they were in a Wes Anderson film, a character in Skyrim, a charater from a 90s cartoon series, and so on. This offers way more flexibility and power than just copy and pasting heads in an image editor. If you check civitai, you will see that creating people, characters from anime and cartoons are one of the most popular categories. There is no denying that there is strong interest in this functionality. Even in the Daz ecosystem, this demand is there. Check out how many posts and characters there are in https://www.daz3d.com/forums/discussion/638911/celebrity-look-a-likes-for-3d-figures-part-4 ;
...well put.
fast food
rendered yesterday on my PC using Automatic 1111
video runtime almost 10 mintes
gormet meal
rendered today using iray
video 28 seconds
I actually spend more time doing 3D believe it or not, just get a lot less finished video sadly
ideally I would like an ethical hybrid 3D and Ai workflow
am still hoping one day it is possible
...for skin tones I've been working with Zev0's skin Builder since G2. Yes there was a learning curve at first, very new programme or process has one, but now it's pretty much my go to for skins that aren't commonly available (like very fair or albino both oaf which I have been very successful with). Many characters with HD morphs also have a separate "HD slider" I can ignore so they work better for combining with others, even older generations. Is I've mentioned elsewhere I rarely if ever use a character right out of the box, save for A3 and Gen4 (even then the V4/A4/S4/G4 "Unimesh" concept that preceded Genesis allowed for a bit more, albeit limited, custom work and clothing fits).
I tend to purchase more Merchant resource and utility content (such as morphing, shaping, and shaders) than characters. as I like developing my own My modelling skills are nothing to write home about and I am unable to use a tablet (which is pretty much essential for sculpting) due to severe arthritis, so I've taken to working totally with morph, shape, and expression controls to make up for that (even a small fraction of a slider input morph can make a difference).
Crikey I was creating teen and even child characters off V4 and later S4.(check out my "Leela" sub-gallery on DA) so I'm no stranger to putting in the time to learn how to push the Daz programme to and past its limits. I'm retired now, I have all the time I need on my hands to work with and perfect techniques using the tools I have and know.
Besides I find it rewarding and relaxing.
Again I feel AI could be useful to improve certain processes like rendering, noise reduction, animation, cloth simulation, and such, but that is as far as I would ever consider employing it.