|
Rank: Advanced Member
Joined: 9/15/2012(UTC) Posts: 426
Thanks: 98 times Was thanked: 156 time(s) in 95 post(s)
|
Hi All! I've been away these last 9 months at Ai boot camp. I know I haven't been active here in the community, but just wanted to say what an amazing this this ai is. I've done a little bit of playing around with feeding the ai my poser images, though, not nearly as much as I should be. I cannot unsee the videos and technology that is already here. Soon (or even now) one can make a 3d figure from a 2d image automatically. Very soon, there will be no need to rig a 3d figure at all, and one can film a quick movie and combine it with the 3d model to generate a 3d movie. I know this seems "off topic" for poser/daz/etc... but if the real goal of a wonderful community like this is to realize our dreams into the world then this tech cannot be ignored. I would love nothing more than to sell tutorials on prompting and using the newest models, loras, embeddings, and hypernetworks. I made this for a client. Not the image...... I was given a data set for the ball gag harness. I then made a set of loras to use this ball gag harness on whoever. I'm not sure how many people see me over on DA or my YT, you might already be aware on how far this has come in a short amount of time. It would be wonderful to hear from anyone in this community about these subjects. Are you all ready for this wave?
|
3 users thanked Darkseal for this useful post.
|
|
|
Rank: Advanced Member
Joined: 4/17/2021(UTC) Posts: 56
Thanks: 16 times Was thanked: 18 time(s) in 15 post(s)
|
Not trying to be a contrarian or to sound like a luddite, but I have zero interest in AI art. Which is actually kind of surprising to me, consider how awesome some of the images are. I don't know what it is, but I just can't work up any enthusiasm for the genre.
|
3 users thanked Katsuyaki for this useful post.
|
|
|
Rank: Advanced Member
Joined: 9/23/2020(UTC) Posts: 344
Thanks: 77 times Was thanked: 187 time(s) in 171 post(s)
|
@Darkseal:
There are some great AI pictures here in the gallery. And if I have the feeling that they were made with heart, soul and intuition, they get a Like.
But I don't think AI will completely replace 3D. Sure - there are quite a few who are only concerned with the production of images. But there are also those who like to work on a 3D scene until all the poses are right, from the eyes to the little toe, and the lighting is exactly as imagined.
This (poses, lighting, etc) cannot be controlled with AI down to the smallest detail - yet.
As a 3D content creator, I personally love pushing the limits of DAZ ... and going beyond. What is possible e.g. with my slipped or take off 3D garments - i.e. exact positioning at a well thought out pose - is not possible with AI - yet.
I have played around with AI a bit in the past, but for me personally it is nothing, because I have too little control - yet.
In my opinion, neither work environment will replace the other. Instead, there will be a juxtaposition - and possibly even a coexistence - and we will see many new images (render and AI): some good, some mainstream ... and some ingenious, made with heart, soul and intuition.
Have a great day!
|
3 users thanked Pushee-Ri for this useful post.
|
|
|
Rank: Advanced Member
Joined: 12/7/2015(UTC) Posts: 83
Thanks: 63 times Was thanked: 47 time(s) in 36 post(s)
|
Originally Posted by: rollemops @Darkseal:
There are some great AI pictures here in the gallery. And if I have the feeling that they were made with heart, soul and intuition, they get a Like.
But I don't think AI will completely replace 3D. Sure - there are quite a few who are only concerned with the production of images. But there are also those who like to work on a 3D scene until all the poses are right, from the eyes to the little toe, and the lighting is exactly as imagined.
%100 agree. What Stable Diffusion does fantastically is render beautifully, with a kind of stylistic coherence.
What it doesn't do so well is compositional IQ, and mechanical accuracy. When I do science fiction imagery -- its easy to get what in the 3D world we called "nurnies and greebles" -- all the random detailing that goes onto a spaceship model, for example. Back in the day of physical modelling, the modellers would often use random bits of plastic, the sprues and supports, to glue on complexity. AI does a great job of making your cyborg "intricately detailed". What it doesn't do so easily is a logically designed functional mechanism.
So, for example, I love Davo's intricate mechanical creations - they're gothic genius. They are going to be hard to replicate in AI - if I want some poor Posette stretched out on the rack, its going to be much easier to pose it that way and import into SD for a "rende"r (it isn't really a render, but if I call it a "generation", that's confusing too) , than it is to actually figure out the prompt that would get you there.
What the 3D content creators have -- is logic and inspiration. If you've made something that _does something_ that's rigged, that's got some story and logic, that's going to be useful no matter how the technology changes.
Just thinking of the work of Nursoda - I've been buy their stuff for years, its inspired, its different than just a texture map, its a complete unique character. I'd I wanted their vibe of troll or dwarf, well Nursoda's actually invented something that I couldn't think of myself.
And here on 'Rotica, there are some incredibly inspired and creative artists too
My advice to creators would be to at least take a look at AI, and think about how you might incorporate it into your sales.
People are making money with SD models -- and there's no reason that folks who've invested a lot of time and effort in building up these assets shouldn't get some return on their investment in these new markets.
|
2 users thanked topspin for this useful post.
|
|
|
Rank: Advanced Member
Joined: 9/15/2012(UTC) Posts: 426
Thanks: 98 times Was thanked: 156 time(s) in 95 post(s)
|
If I wanted to pose exactly the figure I could, right down to the fingers and toes with controlnet. We are now literally waiting for the video models to come out to open source. I expect we see it in the next 6 months that I (hope) to be able to make video. I already have some stable diffusion img2img videos on youtube. There is a funny one that came out today on reddit made from the new video tech with Darth Vader shopping at wallmart. It's crazy. I hope to see you all on the other side!
Edited by user Tuesday, March 21, 2023 11:45:57 PM(UTC)
| Reason: Not specified
|
|
|
|
Rank: Advanced Member
Joined: 9/15/2012(UTC) Posts: 426
Thanks: 98 times Was thanked: 156 time(s) in 95 post(s)
|
tried to embed a youtube video, dont think it worked.
https://youtu.be/j4AIn65RTx8
|
|
|
|
Rank: Advanced Member
Joined: 9/15/2012(UTC) Posts: 426
Thanks: 98 times Was thanked: 156 time(s) in 95 post(s)
|
Ok, to keep Ai out of the image galleries I'm going to spam this thread if I feel the need to. Let me know if there is anything that should ever need to be removed. I'll post some shitty shit and sime less shitty shit. Feel free to comment! Darkseal attached the following image(s):
|
|
|
|
Rank: Advanced Member
Joined: 9/15/2012(UTC) Posts: 426
Thanks: 98 times Was thanked: 156 time(s) in 95 post(s)
|
Some more random images I though no one has seen here. *deleted* Edited by moderator Wednesday, March 22, 2023 12:13:20 PM(UTC)
| Reason: Community Policy Violation - Links and/or Site References Darkseal attached the following image(s):
|
|
|
|
Rank: Advanced Member
Joined: 9/15/2012(UTC) Posts: 426
Thanks: 98 times Was thanked: 156 time(s) in 95 post(s)
|
and a few more, for science. Darkseal attached the following image(s):
|
|
|
|
Rank: Advanced Member
Joined: 9/15/2012(UTC) Posts: 426
Thanks: 98 times Was thanked: 156 time(s) in 95 post(s)
|
this is only the tiniest bit of the shittiest stuff. I dont want to violate tos, so here is more sub par subjects... i might throw in real old crap too
Darkseal attached the following image(s):
|
|
|
|
Rank: Advanced Member
Joined: 9/15/2012(UTC) Posts: 426
Thanks: 98 times Was thanked: 156 time(s) in 95 post(s)
|
If non of that made you think "i'd like to do that" then maybe some o these: Darkseal attached the following image(s):
|
|
|
|
Rank: Advanced Member
Joined: 9/15/2012(UTC) Posts: 426
Thanks: 98 times Was thanked: 156 time(s) in 95 post(s)
|
If none of that made you think it's cool then more of the same won't do it.
Darkseal attached the following image(s):
|
|
|
|
Rank: Advanced Member
Joined: 9/15/2012(UTC) Posts: 426
Thanks: 98 times Was thanked: 156 time(s) in 95 post(s)
|
I do have a few that I did do work on afterwards. Some I did take the time to enlarge.
Darkseal attached the following image(s):
|
|
|
|
Rank: Advanced Member
Joined: 9/15/2012(UTC) Posts: 426
Thanks: 98 times Was thanked: 156 time(s) in 95 post(s)
|
those last ones because they are cute as fuck
|
|
|
|
Rank: Advanced Member
Joined: 9/15/2012(UTC) Posts: 426
Thanks: 98 times Was thanked: 156 time(s) in 95 post(s)
|
Now, let's look at some stuff from 9 months ago... the HORROR!!!! Darkseal attached the following image(s):
|
|
|
|
Rank: Advanced Member
Joined: 9/15/2012(UTC) Posts: 426
Thanks: 98 times Was thanked: 156 time(s) in 95 post(s)
|
I really don't know about tos and celebrities here... else I would have much more to show and not nit pick. But like, this seems to be working pretty well. I will be running some poser pics through in the future, but I've been having plenty of fun so far not really needing to yet. When I do give it something, it's usally video.
Edited by moderator Wednesday, March 22, 2023 12:22:46 PM(UTC)
| Reason: Not specified File Attachment(s): Darkseal attached the following image(s):
|
|
|
|
Rank: Advanced Member
Joined: 8/1/2013(UTC) Posts: 89
Thanks: 133 times Was thanked: 69 time(s) in 65 post(s)
|
Hi Mr DarkSeal, long time no see.
I haven't been here in over a year, just popped in to buy Davo sale stuff (New-ish electro stuff looks awesome).
- Saw your post. Visited your *deleted*. THANK YOU for reinvigorating me in 'art'.
- Setup the Kohya_ss gui and GIT and python (well... I had some help...)
- Watched the guy's 2 training videos and realized I needed SD too.
- Started on that, currently downloading the ckpt file now (8gb, seems to be quite slow...)
- Going to start looking for some Angelina Jolie to start...
I'd love to know how you achieved the slime/cum lady-tentacle effects, and what you used to train.
Going to try the blip captioning, but I'm a bit confused. Comments and google tell me I should caption everything. They also tell me I should only caption specific things.
Any advice is greatly appreciated, as it pointing me at any posts/tutorials that helped you.
All I want to make is some larger ladies that aren't pulling a stupid face or have hairy crotch, and some muscular girls who don't look take steroids and don't have giant clits. Too much to ask for? Edited by moderator Wednesday, March 22, 2023 7:32:42 PM(UTC)
| Reason: Not specified
|
1 user thanked Floor13 for this useful post.
|
|
|
Rank: Advanced Member
Joined: 9/15/2012(UTC) Posts: 426
Thanks: 98 times Was thanked: 156 time(s) in 95 post(s)
|
Hey Floor13!!! It's been awhile! Glad to here you get koyha set up, but that's for training only. You want to set up Automatic1111 on github, as well as python 3.10 and git for github. THose 3 things make it all work. After that you'll use different models, embeddings, loras, vaes, and hypernetworks to create a symphony. Start here: https://github.com/AUTOM...1/stable-diffusion-webuimore specifically here: https://github.com/AUTOM...installation-and-runningThere are so many models that do so many specific things... but that tentical is one of the ones I photoshopped! The face was "suprised" i think in the prompt. Sometimes it's just that easy. 5-10 min. in PS after making a bunch of renders to source from! As for "what models to get"... MINE OF COURSE! I've got a bunch up for download, all free atm: https://civitai.com/user/DarksealOh, BTW, I just got hit with some kind of underage warning from renderotica here. If I get auto kicked It will be for stuff I've posted in the last day or 2. As for the girls... You can make them any shape you want floor13. Darkseal attached the following image(s):
|
|
|
|
Rank: Advanced Member
Joined: 8/1/2013(UTC) Posts: 89
Thanks: 133 times Was thanked: 69 time(s) in 65 post(s)
|
NEVER had by graphics card spin like that, even with Cyberpunk at 4k 60FPS raytraced.
RTX3080, ran out of memory once trying to high-res.
Well, this was a quick attempt, apparently can't do nude...
Doesn't look anything like her, she was at 0.7 strength.
Floor13 attached the following image(s):
|
1 user thanked Floor13 for this useful post.
|
|
|
Rank: Advanced Member
Joined: 9/15/2012(UTC) Posts: 426
Thanks: 98 times Was thanked: 156 time(s) in 95 post(s)
|
Originally Posted by: Floor13
RTX3080, ran out of memory once trying to high-res.
Well, this was a quick attempt, apparently can't do nude...
Doesn't look anything like her, she was at 0.7 strength.
It might be the model/lora was bad and has clothing. Just try my Pinocchio 1.6, she's in there.
As for your memory issues, you can add -medvram to your webui-user.bat file. Mine currently looks like this:
set PYTHON=
set GIT=
set VENV_DIR=
set PYTORCH_CUDA_ALLOC_CONF=max_split_size_mb:64
set COMMANDLINE_ARGS=--theme dark --no-half --opt-sdp-no-mem-attention --xformers --medvram
set OPTIMIZED_TURBO=true
set ATTN_PRECISION=fp16
set SAFETENSORS_FAST_GPU=1
call webui.bat
................................there is also a extension called mutidiffusion upscaler: https://github.com/pkuli... but with medvram things will be slower but finish rendering. I know that upscaler is godlike, I have not used it to its full potential.
Darkseal attached the following image(s):
|
|
|
|
Forum Jump
You cannot post new topics in this forum.
You cannot reply to topics in this forum.
You cannot delete your posts in this forum.
You cannot edit your posts in this forum.
You cannot create polls in this forum.
You cannot vote in polls in this forum.
|