AI is going to be our biggest game changer
This discussion has been closed.
Daz 3D is part of 
Connect
DAZ Productions, Inc.
224 S 200 W, Suite #250
Salt Lake City, UT 84101
Licensing Agreement | Terms of Service | Privacy Policy | EULA
© 2024 Daz Productions Inc. All Rights Reserved.
Comments
more
are too many
A reminder.
Do you do this, too? Just generate a bunch of images on a theme, and then just stare at them letting the neural in between your ears process them, and come up with an idea inpired by them but distinct? And sometimes it is actually an imperfection in the image that breaks you out of the mold, and you think of something truly creative that you would not have thought of, had the image been pristine.
As the images get better, as they markedly did from v2 to v3, we'll go losing this along the way. I hope there's a setting to inject certain degrees of surreality(sp?) just as a seed for creativity because sometimes you don't know what you're looking for until you find it.
most of mine are flukes, it never gives me the result I expect.
but that is what triggers a primal gambling like feeling which is addictive
it is taking a lot of willpower not to raid my main bank account and subscibe at the $30 /month rate
I would need a multi terabyte drive too
My imagination is more triggered by seeing others results
but 3D art also does that
I was actually looking for a DAZ resource for this image, but wasn't quite getting there with the ones I found so tried my idea in MidJourney and was blown away. There were several iterations and upscales and variations, and I ended up compositing elements from other iterations and in the end was going to use DAZ to at least render some more human looking eyes, but ended up taking a shot in Photoshop's Neural A.I. to "change the gaze" and it sorted some pretty wonky eyes right out. These are the results. I LOVE *ALL* of this stuff!!!
MidJourney prompt:
/imagine Frankenstein's monster in a three piece suit, looking important. --version 3 --h 1920 --w 1200
Photoshop Neural AI corrected eyes"
MidJourney composite of several MidJourney images:
I did take one of my original renders:
And fed it to the AI:
I caved and resubbed
https://www.facebook.com/wendy5/posts/pfbid02hpfgtV329x4PaDwDvpbvvuFDvtK1MN6GaZ6kVJEnGPyCzRxxgTo4uxPdTLzzUdg1l
welcome to Elvendale
I got the basic sub. Though it really isn't enough images a month. I hope they will bump the monthly number up.
But still cheaper than my monthly DAZ budget used to be by far. :D
All three of the art cllections you posted are visually stunning.
Quoted for complete agreement
Nothing I do comes out even half as vibrantly detailed as that, Wendy. If certain people can still be better at it than others, perhaps not all of the artistry is gone yet. Please share what your "knack" is :)
The Creation of Adam painting. That's crazy.
I combined 16 upscales to get this
I cannot take credit for any of it BTW and have seem far more brilliant stuff on Facebook other's generated
Kind of like how we aren't artists because we just combine pre-made assets.
Knowledge of command line options? And an idea what ingredients night lead to a nice result, like when a cook is trying to design a new meal, a brewer a new beer etc.
I started playing with Midjourney over the weekend. So far I'm on the fence. Yes, it can produce some wonderful stuff...but nothing I've seen so far ends up looking...real. Like when you look closely, everything is all jumbled, like an amalgamation of so many image pieces layered on top of one another to make up a whole. It could look great if that's what one is going for, but it's not my cup of tea. For creating trippy backgrounds and such, it could be a great tool...when combined with a rendered foreground or portrait.
Another thing I'm not a fan of, at least with the free trial images, is that someone can see your results and then copy your prompts ver batim and create something extremely similar. This happened to me when I put prompts in to get this -
I guess if/when one starts to pay, they can work diretly with the bot via dm.
Lastly...there's no "signature look". Like I can look at a lot of artistic renders and pinpoint who did it based on the style. I know I even have my own signature look. But images created via AI all seem to have that same schmish schmash look, no matter who dumped the prompts in.
Stable Diffusion is now in beta and has some big differences over Midjourney and Dalle-2. Stable Diffusion is being developed as open source and will eventually be made availble to the public to run on your own hardware (although I believe it will require a pretty beefy Nvidia card to run). Stable Diffusion also will not be filtered, unlike Midjourney and Dalle-2 that don't allow things like certain people or subjects.
I think Stable Diffusion does a really good job of making photorealistic images of celebrities, but more melty/distorted faces when trying to make generic people. Stable Diffusion also doesn't seem to follow the prompts as closely as Midjourney and Dalle-2 but hopefully it will improve over time as they further tune the model.
Attached are a few pictures I made and a few I found in the SD discord server showing some Daz style portraits. These took only seconds to generate which is crazy thinking how many hours it would take to craft these in Daz Studio (including the ten minutes of snooze time as Genesis 8 loads zzzzzz...).
I got an invite to DALL E a few days ago and have only just now started to have some luck with it. I feel like I wasted a lot of my free credits (I guess that's to be expected) figuring it out haha. Apparently "realism" or "realistic" doesn't help get realistic results. "Photo" worked for the most part. I still had a few images that looked "painted." It would probably help if I uploaded some photos for the AI to work around. I liked how this one came out though. I used the prompt "photo of a shadow lord standing in a forest at twilight." I mean, not exactly what I was thinking but it works.
Imagine a DAZ3D render:
well DAZ do own Bryce
How wonderfully meta of you :)
So that's what we've all been doing. But where's VWAS?
:D :D :D I was just thinking that very thing. A true test of AI,,, though if it can do it? I fear it might be sentient.
Or the user who set up the command line was a computer genius, using exactly the commands needed to achieve such a result.
Still need something like DAZ. As for now, he AI can't do good looking limbs, and it is terrible with hands and feet. Not possible to produce dragons with a real functioning body or horses, let alone a horse with a rider.
But it is brilliant for landscapes, all things ornamental and portraits.
And no taking pictures from another angle...
I jumped back in WOMBO Dream last night. One thing I like about Dream is there's a lot of styles to choose from (The Ghibli one is pretty cool). I prefer the dark fantasy one. Some of it reminds me a bit of the Guild Wars 2 art style (which I LOVE, so this was appealing to me). I will say that it is sometimes hard for me to distinguish this type of AI generated art from something a human painted. I've seen so much art like this in the 18 years I've been on DA. There has been a few times lately that I've seen something similar pop up and wonder if it was painted by hand or generated by an AI.
the Midjourney hands
https://www.facebook.com/wendy5/posts/pfbid0s7jSVV9M462GRUe9bfrYgzaqz53Z7auw3hP6WnjrQqNhHD5fjZc35DNvzb1Y8QSql
https://www.facebook.com/wendy5/posts/pfbid02iGmhzLFm52wFQ5UMpMrBGGTsgCTFC47X9jf4uiFBqjFgAdvgvwa8VZ5mg11VYw7Bl
it is horrifying
I rendered a bunch of DAZ hands to Gimp postwork in
/imagine
There is another called Nighcafe. Its poretty cool and seems less expensive than MidJourney. They give you credits for a lot of stuff. my work very elementary yet, its text based and there are multitudes of modifiers.
I have seen some images that just really blew me away ...
Here are a couple very beginner ones I've done,
Even this bad stuff is entertaining. Describe something and see what gets created. The blue one I typed in Crystal Blue Persausian..an old song title. The other I just typed in truck asd HR Giger style.