DAZ 3D and the new Photoshop Neural AI

I just found out about Adobe's update of Photoshop to include Neural A.I. filter effects. Overall I am impressed, but also a little disappointed. Some items are clearly marked BETA and so Photoshop's image FX meant to be like Google's Deep Dream Generator fall far short, but some of the Aging, and other effects are pretty amazing. When combined with 3D they can literally make or break an image, or merely give you more options. I know some 3D Purists are going to HATE this, but for me, "Whatever tool gets the job done" is what floats my boat. I'm ALL about the final image.
Here are some SUPER quick, unedited examples of the new Neural A.I. effects from a Photoshop update download that hadn't been on my PC long enough to let the paint dry. The original image in these examples is a Freja 8 render I did and it was post-processed. I just grabbed the first image I found and ran with it. See examples for an idea what to expect and what is possible including running multiple effects on the same image in the same filter session - Niiiice. I understand MUCH more about it just from doing this sample series and it is pretty easy and intuitive to use.
Note: There are TWO versions of the Happiness effect to show that you may get different results running the same effect two different times on the same source image. AND of serious note is that those teeth are 100% extrapolated! No teeth were showing in the render at all. I've seen better teeth, but also these teeth do not jump out and kick me in the .. uh... shin.
Here we go - INTO THE FUTURE, AND BEYOND!
















Comments
Wow - those are pretty impressive for something that's still in a beta stage. Although I can see how some may hate any automation to the creative process, I think the possibilities for those of us (read: me) without a surplus of creative talent will be endless. I can totally see myself getting lost in creating variations of an original concept.
I noticed in the gaze example the eye colour has changed, which means it has not used the original eye part of the image to create it.
Well, it has to be inserting data from elsewhere or it wouldn't be able to magic up teeth.
I got the exact same looking teeth when I tried it, so it's probably using a set of teeth (and eyes) in the cloud to generate them. I'm really impressed by these plugins though, the head-turning one is really something when used in very small amounts. Cool stuff.
Looks intersting and potentially useful. I hope the method gets taken up by someone who lets you buy the software instead of paying for it over and over again.
I tried it with some photography, and it wasnt very good... but i think it's still an easy-to-use tool so for that im glad it's there.
my intended use-cases were re-lighting, posing, and changing expressions.
I found that all of these at best were average, and at worst were very bad.
I imagine people will fix those old unsmiling family photos with it
Mostly it's really really creepy.
https://youtu.be/hq8DgpgtSQQ
It's like everything, though. A little goes a long way. And dangerous in the hands of children.
I used it with some non-photorealistic, graphic styles and it works pretty dang good. As a beta this is impressive.
I assume, seeing what site you are on, you're being ironic about it being "creepy"?
That's really neat. It definitely needs work though. The fifth and sixth images (the ones with smiles) in particular - the mouth and teeth are quite wonky.
The aged one kind of works, but they gave her double eyebrows and the neck is really funky.
To me, the most successful one would be "Surprise".
It's interesting the direction it's going with AI "assistance" in art programs. Though, for good or bad, we seem to be getting closer and closer to that, until now, completely fictional "make art button".
I purposely posted these "as is" from the Neural filters, but with just a little smoothing and Photoshoppery I think they could look much better! The good news is "Things are happening!" and at least for now these fx are still optional. I think they will very quickly become extremely ubiquitous now that they reside in the hands of the masses though, for better or worse.
One thing that I personally thought was great when I was watching a presentation about the new PS features was the pattern preview. Actually, there's tons of good stuff in the new PS.
You can't trust what you see in the media now since the professionals have access to advanced photo-manipulation and deep fakes. Soon you won't be able to trust your old family photos either. Isn't technology wonderful.
I do genealogy and from between 1850 - 1930 you can see photos of absolutely unsmiling faces, especially those folk that were sustenance farming in rural areas. Women and girls especially. When I see them, first thought is always, "Oh boy, I hope they were a lot happier than they looked!" I wouldn't change the expressions though as they might not be indicative of actual mood but there was a school of thought among some folk that one shouldn't smile for the camera.
In those days photography required long exposures. RBF is a lot easier to maintain than a smile, which would look forced if held too long.
This is so fun! The head rotate thing I can also see being useful where you have an almost-straight on photo for FaceGen but not quite.
Can't wait for it to get to GIMP, LOL. I cannot be bothered with an Adobe subscription.
There is more and more AI stuff out there, including FOSS stuff on github... like AI that does interpolation to add more FPS to your animation work, for instance.
It's an amazing time for Skynet to murder us all.
I've played a bit with DaVinci Resolve (free version) and I'm pretty sure that it also does the interpolation adjustment necessary to change your frame rate (up or down).
As for this Photoshop trick ... I think I'd rather stick with using the tools in DAZ Studio rather than attempt it in post. Not that I would ever subscribe to Photoshop anyway - I'm quite happy with Affinity Photo for most things and Gimp for things that AP doesn't do.
They're not going to get very far with only 10 gigs of video ram.
Yep! As soon as the process sped up and cameras became more affordable, the formality eased up and people started taking more casual photos. Vintage selfies from around the turn of the century are really fun, and artists started experimenting with the creative applications. I love looking back at when technology we currently take for granted was avant-garde and people were thinking of lots of different uses for it.
I tried the neural filters on a few renders and so far I'm most interested in the style transfer (although it crashes a lot). Especially with the ability to preserve color, it might be a good way to create lineart solely from filters. The portrait stuff hasn't worked out for me yet, but I'm curious to see how it evolves.
I have Photoshop CC new update .I found two neural filters : skin smoothing , style transfert
How or where can i have the others tou use
Thanks
There's what looks like a little Erlenmeyer flask I think right below those two options, and if you click that it takes you to the beta filters.
Oh, that's makes sense.
FWIW I was hoping to run these FX on this image of The Child and Photoshop wouldn't even activate the filter for it - didn't recognize it as a face, but worked right away when I tried a different image, "Hikaru Redd Blayde" with Neural Face Alteration). Interesting. It seems to be "pre-checking" an image before even offering to attempt to filter it. I wish it would just let me TRY.


Photo by Erik Pedersen, aka Gogger, of The Child semi-plush figure and WIP 3D render by Erik Pedersen.
And this should be it for a while, but wanted to show something less freakish or uncanny-valley as a way of using Adobe's Neural AI effects for more subtle, usual FX.



3D render of Hikaru with Monica Hair in DAZ Studio then turned to B&W using B&W Image Toning in Filter Forge to prepare for the final image, Neural Colorized in Photoshop 2021. I can see me using the full spectrum of Neural A.I. effects in both blatant and subtle ways. Two opposable, digital thumbs up!