Daz AI Studio - Official Beta Launch!
DAZ_Steve
Posts: 42
Want to learn more? Check it out:
Blog: https://blog.daz3d.com/introducing-daz-ai-studio/
Youtube: https://youtu.be/USA5pD3Q2jA?si=KpS6pVt5XXYDGdP3
PR: https://www.prnewswire.com/news-releases/daz-3d-unveils-daz-ai-studio-with-partner-stability-ai-302106757.html?tc=eml_cleartime
Guide for effective prompts: https://blog.daz3d.com/daz-ai-studio-prompt-engineering-guide/
Post edited by DAZ_Steve on
Comments
It worked for me. It just gave no indication that it was doing anything, and it took a long time.
I hope this is a delayed april fools joke
it s already hard enough to get into event art alleys explaining the difference of 3D render and AI to tech noobs with the AI ban sledgehammer swinging organizers. Now it will be almost impossible. Thx in advance
Doesn't work. First image I did I got 4 arms ..
It's not working.
I'm also hoping this is a delayed April Fool's joke.
I'm also wondering how those lawsuits against stability ai are going.
all i can do is choose V9 or M9, nothing else. Cannot edit the created image
So, I tried a couple of test prompts for the giggles. I am not seeing this as an asset in any way, shape, or form. Anything beyond a basic prompt of Woman Smiling ends with deformed eyes, hands, other limbs, etc. and leaves you with an image that needs more postwork than you'd ever want. Mix that with the generation time and unsure results and there is absolutely no benefit to adding this to my plate. I'm sorry, but I've already spent plenty of my content for Daz. Why on Earth would I be interested in sketchy results at best when I can create a sure scene?
Yes. Let's hope this is a late April Fools joke.
Keep in mind this is a Beta so still being worked on.
I was hoping the AI bit would be added to DS to let you use prompts to set up environments and lighting and pose characters you already own--basically use AI to make DS easier and posing less fiddly.
Having a whole 'nother image generator is a real letdown for me.
Is it too much to hope that it will just be cancelled? If it's in beta then hopefully you haven't spent too much time or money on it and you can just walk away without much loss.
It literally cost YOU nothing, so why would you want it to be "cancelled"? If you don't want to use it, don't use it. Those who want to use it can use it. When did we become a culture of "If *I*, personally, don't want to use something it shouldn't even exist".
I for one think it's a nice option for people who do want to use it. :)
Semi-interesting, but I don't see a use for it - and my first two test runs generated V9 with six fingers (well, five visable fingers and presumeably a hidden thumb).
This need to be merged into daz3D and use our own (made) and owned characters.
It can then use to set up a scene with chartcher and poses where you just need to make minor adjustments. Then also render 3D with ai enhancement. I hope that's where this is going or I will be very disappointed.
The text to 3D tool is still being worked on.
Unimpressed, hope you can achieve something, which works with our daz studio library asset's. I'm a big midjourney user, and can get really good renders out of it. In order to stand out, the idea to use our library assets to build an ai scene, genius if you can pull it off. But be aware I'm not looking to Re-purchase upgraded assets to get this to work for you.
So far I've been unimpressed with daz studio bug filled releases, so you need a better standard for ai. And a big purchase on gpu renderfarms to get the render times down to a minute with the ai, to match the speed of others.
price wise it will need to be competitive in the $30 per month for fifteen hrs of render time. Otherwise at the rate that the other ai systems are progressing, you will be the least attractive offering out there.
https://blog.daz3d.com/introducing-daz-ai-studio/
First and foremost, I would love to know what content this generative model was trained on. Given that this is one of the main controversial aspects of generative AI I think it's in the interest of everyone involved to have maximum transparency here, so everyone can make an informed decision! I think there's no better way for DAZ to show their stated commitment to make generative AI be a positive for artists!
Beyond that, I did play around with the free version a bit and I have to say I'm not impressed with the results so far, but I will also concede that I only spent less than 30min with it, plus it's still in beta, so will keep an open mind! Would love to see some videos showing us specific and concrete examples of creators working with this generative model, showing us concrete workflows, showing us how to make it do things that competitors (eg midjourney) can't do. The promotional material so far is heavy on buzz and light on specifics, so I think more specifics will be the best way to sell us this new development!
I'm also low-key confused how exactly this generative AI model relates to DAZ as we know it and the many products in its eco system. So more clarity in that aspect would also be appreciated!
You say it's a LORA of only daz owned images, but which base model are you using it on top of?
From a swift walking over the prompt, i can say, that it tends to make multiple objects of my prompt, which has me trying to describe a space freighter colliding with the side of a mountain after takeoff. then dust, sparkles, rather a cargo spaceship. With freighter it tends to give you something on a trailer, maybe an extra truck, or up to a space freighter, a helicopter above and then a sports car driving the other way (...). So maybe there is a better way of addressing what one object/thing has for specifics. Prompts need some trying, and there probably isn't a generic way that can be learned. You might have to fiddle with each specific tool in a way specific to that tool, in order to find out, how to steer it. (Maybe, worse, for each training round or rather major version change.) So i would try some angles, before dismissing a tool of this kind.
@TheKD LORA? Where's that information from?
From the thread in the commons I believe
An AI assistance to enhance our own renders by making suggestions or suggesting better ways to render subjects would have been helpful. Or taking info only from your own library of works that you own... would have been great too. But not fom other artists works. I feel uncomfortable as this looks to be compromising my principles. Thank you, but hope it does do more than just using other people's artform.
OK, yeah, https://www.daz3d.com/forums/discussion/680756/daz-ai-studio-discussions/p2 Richard Haseltine said it's a lora, and the press release https://www.prnewswire.com/news-releases/daz-3d-unveils-daz-ai-studio-with-partner-stability-ai-302106757.html?tc=eml_cleartime said it's trained on SDXL from SAI. So I guess it's an SDXL lora that's generating the images.
Are you friggin' kidding me ?!?!?! This is just so wrong, let me count the ways!
It's been months since I opened DAZ Studio because I've been addicted to creating AI artwork (Stable Diffusion -> CivitAI-> Automatic A1111 -> Totally free, with hundres of more options).
I finally said it's time to get back into creating real art and get back to DAZ Studio. Only to find THIS... A very limited, poor man's attempt at AI artwork. (even with the paid subscription).... not even close to what is possible with a free offline AI program (Stable Diffusion - WebUI)
Call me when your AI program can actually prompt images to work with (3D) Daz Studio. I'll be back when it does.
This is just a slap in the face to AI enthusiasts and especially long time DAZ customers!
You should have STARTED with the "text to 3D tool" for DAZ Studio. Until that time this thing is virtually useless and a huge insult to long time, loyal DAZ customers!
not really impressed by the results myself
needs ControlNet to be useful IMO
I use my renders in Controlnet for Stable Diffusion on my PC with openpose, depth and canny to get consistant results
While the first AI images looked stunning, because of their status of being new, the look of AI has become old.
You could say AI images all look alike. Over saturated and boring.
Thanks for creating it. I am impressed by the results so far.
I have created a separate thread with my experiments with AI generated images.
https://www.daz3d.com/forums/discussion/681031/artini-s-aiteracy#latest