- This topic has 11 replies, 9 voices, and was last updated 3 years, 4 months ago by
.
-
Posts
-
Give it another year or two, and DAZ may be completely obsolete: https://huggingface.co/papers/2305.10973
Video fashion change and pose demo: https://s3.amazonaws.com/moonup/production/uploads/60a551a34ecc5d054c8ad93e/asSF_QiOtZ-Iqv2fV2-QE.mp4
Project page: https://vcai.mpi-inf.mpg.de/projects/DragGAN/
Honestly I'm not a artist in any way but everything Ive seen and done with ai art seems very randomized and can take you any amount of time to get close to even 70% of what you want. I'm pretty sure a artist could make something much closer to what you want in less time so I dont think artist are going to be going away.
I don't see artists going anywhere. The AI is only generating stuff based on what it's already seen or been trained on. It's all random variations based on a theme or original picture. The AI can't be creative or original.
I'm not sure that 3D will disappear either. That video of the clothes being altered in real time is nothing that Marvellous Designer can't already do. I can see AI being incorporated into say Blender to significantly speed up item creation and of course, rendering.
Daz will go under because the can't keep up. Studio 5, if it ever appears, will be similar to Poser 13, already woefully out of date and badly written. Meanwhile we'll be playing with our massive stash of content in whatever comes next, maybe one of the game engines?
@Veck17 "...ai art seems very randomized and can take you any amount of time to get close to even 70% of what you want."
Which means it's virtually useless for VNs as there's no way to reproduce consistent characters or scenes. One click may get you something you want, but the next one will give you a character with arms growing out of their ears in a completely different environment.For what it's worth DragGAN just went up on Civitai.
Call me old fashioned if you like....
I do not even want to experiment with AI. What I have seen so far is bland, perfect but bland.
I heard Ai generated popmusic ( Oasis kind of sound), AI even did the singing. It sounded like Oasis, and if I was not told it was AI I wouldn't have known.
But is is al to smooth and perfect for me.
Also you would mis the joy when an artwork works out the way you want.
As I spend more time in DS, I am getting better in adjusting the little things ( A hand gesture, an extra ghostlight, some prop I find suiting for the image) Things I guess will stil not work in AI, you tell what you want, you get.
Of course you can try again with new parameters, but the creation does not visually unfold as you are working.
I work with clay as well, and when I sculpt, the sculpture " matures" as I go along.
It is this " growing under your hands" that I find most satisfying about creating things....
My feeling is that AI does have a part to play, but it's not going to kill the creative (3D or otherwise) processes. Instead it will be integrated into the render engines to add more detail, denoise or to allow rendering in specific style. Also for things like animation so for example, you keyframe a character at point A, move them to point B and just tell the AI 'Walk between these points using 'sexy-walk-style-2'. Not to mention its use for things like texture creation and mapping. Imagine if, instead of having tens of 4K and 8K skins textures to cram into your VRAM, the AI renders a few dozen iterations with low res textures, then creates and maps high res ones on the fly and applies them to the high-noise, low-detail render.
I'm kinda on the same viewpoint as Skippy; It would be very nice to be able to set up a scene in Daz (or) whatever comes next, and not have to mess with shaders or anything, just say "render in a 1960s comic book art style" and it produce something that wouldn't feel out of place in a vintage comic collector's stacks for example.
As it stands, AI art is interesting but a long way from being useful for me. VNs and comics require consistent character appearances from one image to the next, and even the best engines I've seen can't manage to pull that off. Good for concept art, shit for consistency.
You have motion capture movements on file. Hundreds of years of art to reference.
3d body scanning. AI voice and what PS can do ATM is pretty amazing. Deep Fake.Avatar 2 is basically nothing but CGI. Most Marvel movies are filmed in a room.
Writer and actor strike? Who cares pop out a CGI movie.
When it comes to me I will do a render in DAZ then run through IMG2IMG and experiment. If I come out with something I like, I will import my render and the IMG2IMG result into photoshop and experiment with filters, extra layers, blending, etc. It can't replace the work I do in DAZ and unless there is a massive leap where I can produce exactly what I want every time it will never replace what I do in DAZ. My next goal once I finish my new desktop build to replace this aging clunker with a Ferrari engine (i5-6600k paired with an RTX 3090) is to train my own model using my renders. Such as making a character and doing multiple renders with cameras in different positions then using the renders to create a model.
- You must be logged in to reply to this topic.
