|
Mozi posted:Yeah it's great, the only part we have to change now is needing to make money in order to live. That's doable. I'd explain how but it's against forum rules and federal law.
|
# ? Dec 2, 2022 18:57 |
|
|
# ? May 22, 2024 17:00 |
|
Neat. Unsolicited tip: To prevent the double heads when rendering at 1024x1024, you can either use the HighRes Fix setting in Automatic1111, or render at the model's native resolution (512x512 for SD 1.x, 768x768 for SD 2.0) and then use the upscaler of your choice.
|
# ? Dec 2, 2022 19:04 |
|
White Light posted:Been looking at some of the things posted in this thread and woof, methinks I'm gonna have to make a hard pivot from my years of art practice, ain't no way I can compete with an algorithm like this, good lord Not sure what you do or what your workflow is, and I don't mean this like it's an easy thing to do, but all of these things being posted are really cool yeah, like 90% of the way there and honestly well past the 'good enough' line for most people especially the general public. HOWEVER all of these AI results can use a bit of finishing, just a person to go over them a little bit, and they'd be 110%. I went to school for art though never worked professionally, and it's very advantageous to have that ability to work with the system to get really good results knowing the language. This stuff is an incredible tool, that's all it is. People have already been posting in here about how they are using it to create stuff they'd normally do it. In my case I'm not sure how you'd create photos of things that can't exist besides a time consuming 3D modeling -> rendering setup for each image. Each one of these can be brought up to another level with just the most minor of tweaking. The biggest draw for this tech is making these ~real~ photos that couldn't exist in any other way without an insane amount of work. As someone that never developed any drawing skill these tools are amazing.
|
# ? Dec 2, 2022 19:15 |
|
|
# ? Dec 2, 2022 19:19 |
|
If anything the feature of taking little line drawings to push things in the right direction and make really cool poo poo just from that has actually made me start to draw a little E: as a lifelong amateur photographer, everything kharmakazy said is very, very true. AI art is just a new art form, not an art killer, despite what everyone on Twitter thinks TheWorldsaStage fucked around with this message at 19:25 on Dec 2, 2022 |
# ? Dec 2, 2022 19:21 |
|
Cabbages and Kings posted:So, tl;dr on that, "same, kinda". I have dabbled with silly creative things here and there, and to me Stablediffusion has just become more of that. Look at this, it's like that old Macy the Mouse book we had at my house ❤️ I dump all my stuff in this one thread every few months
|
# ? Dec 2, 2022 19:34 |
|
inpainted background, astral fields in the style of dali and van gogh, negative prompt fog, midjourney 4 in stable diffusion ui using ddim sampler, everything i did just made this tooth monster that looks exactly the same scaled down MrQwerty fucked around with this message at 20:00 on Dec 2, 2022 |
# ? Dec 2, 2022 19:46 |
|
AARD VARKMAN posted:ok come on Yeah too far, my bad. I'm feeling because of the current empathy for illustrator jobs where it didn't seem to appear for other recent disruption. Music, factory work, even programming itself with Github Copilot. Gotta get out of the crab bucket mentality and focus on the bigger picture
|
# ? Dec 2, 2022 20:01 |
|
Found out that https://github.com/ashawkey/stable-dreamfusion allows one to generate 3D models based off Stable-Diffusion prompts and spent half a day yesterday getting it installed only to find out that I'm very unlikely to be able to run it with my 4GB VRAM. And I'm a couple of significant price drops away from being able to afford 16GB GFX cards. My only hope is that others do some cool stuff with it. Maybe if it gets more attention someone'll figure out how to lower the VRAM requirements sufficiently.
|
# ? Dec 2, 2022 20:08 |
|
Cabbages and Kings posted:So, tl;dr on that, "same, kinda". I have dabbled with silly creative things here and there, and to me Stablediffusion has just become more of that. 👑
|
# ? Dec 2, 2022 20:09 |
|
AARD VARKMAN posted:I've been using my apple pencil a lot more recently drawing stuff to throw at the AI so I can get ideas across that it struggles with. Also been doing ridiculous hack Photoshop jobs using MSPaint. It would definitely be easier if I were actually good at either of those things I am using AI images to teach myself photo retouching. The fact that so many details are hosed up is great for giving you little things to practice on. Fixing up eyes, teeth, fingers and removing the occasional extra limb are great exercises. I am using Affinity Photo though and not MSPaint. Highly recommended if you don't want to spend buckloads of $$$ on Adobe (even if the recently released v2.0 came with a minor price hike, I'm still on v1.x though, good enough for me).
|
# ? Dec 2, 2022 20:13 |
|
i've started looking at the hands in every human made painting i see, and half of them are hosed up. i understand why the AI has problems
|
# ? Dec 2, 2022 20:31 |
|
Someone should make a model on just Rob Liefeld art. Edit: enjoy Cousin Todd fucked around with this message at 20:52 on Dec 2, 2022 |
# ? Dec 2, 2022 20:46 |
|
Cross-posting this since it's very relevant to the thread. I'm doing an illustrated Choose Your Own Adventure version of Jurassic Park in the Let's Play forum, a sort of side-quel that takes place the same day as the first movie. The whole thing is powered by (mostly) AI imagery, and hopefully will be a good bit of fun: Jurassic Park: Edge of Chaos Planning to incorporate AI text prompts, too, but we'll see how chaotic it gets.
|
# ? Dec 2, 2022 20:53 |
|
Did it get better at dinosaurs? Last I checked the ai just made weird monsters instead of dinosaurs by default
|
# ? Dec 2, 2022 20:57 |
|
No. The dinosaurs will be very bad. Some of them will work okay, but there will be a lot of extra limbs.
|
# ? Dec 2, 2022 20:59 |
|
AI art for role-playing is absolutely fantastic. I've been using midjourney to make all kinds of player aids for my D&D game. Here is a bunch of the pictures i generated for our last game, completely out of context
|
# ? Dec 2, 2022 21:03 |
|
Kharmakazy posted:Art isn't a zero sum game. You don't win art. AI won't destroy the ability to create meaningful art for the sake of making art. People are always gonna want to do and want to see that. And people definitely put a premium on 'genuine' that you can't get through AI. But you can't pay your rent with emotion and meaning. It'll certainly hurt job opportunities for people who are paid to make soulless corporate art, stuff that exists to make money or get clicks or whatever rather than to send a message. Like ads, or stock images for clickbait articles or whatever. In 5 years It'll likely be a lot harder for artists to make a living doing art as their day job. That kinda sucks. UBI would be cool. in conclusion, Mozi posted:Yeah it's great, the only part we have to change now is needing to make money in order to live. RPATDO_LAMD fucked around with this message at 21:13 on Dec 2, 2022 |
# ? Dec 2, 2022 21:08 |
|
Kharmakazy posted:Did it get better at dinosaurs? Last I checked the ai just made weird monsters instead of dinosaurs by default feedmyleg posted:No. The dinosaurs will be very bad. Some of them will work okay, but there will be a lot of extra limbs. Since the earlier posts about dinosaurs was with MidJourney I decided to try it out with Stable Diffusion 2.0. Raw text-image outputs with no cherry picking, 1204x768 (besides brachiosaurus where I flipped the ratio) which will account for the prevalence of extra limbs: "dinosaur" "stegosaurus" "triceratops" "ankylosaurus" "brachiosaurus" Considering the prompt is just the one word I think Stable Diffusion 2 is quite good at dinosaurs. With a bit of tweaking I think you could make some pretty radical dinosaurs. If your intent is to make pictures of already-existing ones though it's going to be difficult, a lot easier to make dinosaur-like things. Just claim they are never before seen ones!
|
# ? Dec 2, 2022 21:12 |
|
maybe they need to stop feeding it D&D manuals because i'm pretty sure those are basilisks
|
# ? Dec 2, 2022 21:31 |
|
The results look like rear end and the first test run took an hour and a half but I typed in words and I got a 3D model! https://colab.research.google.com/drive/1MXT3yfOFvO0ooKEfiUUvTKwUkrrlCHpF?usp=sharing (That's the link to do it yourself on Google's servers; first time I think I've been happy to have a Google account.)
|
# ? Dec 2, 2022 21:50 |
|
It was always a bad idea to get into art lol
|
# ? Dec 2, 2022 22:30 |
|
"i'm gonna get into art for money!" - someone who doesn't know human history
|
# ? Dec 2, 2022 22:34 |
|
So I don't tend to post in here because I'm a dirty ole man and most of what I make is women in various states of undress, but the RNG blessed me today when a low-effort prompt produced this: I didn't even have to use inpainting to retouch the eyes, it was like this straight out of the oven.
|
# ? Dec 2, 2022 22:42 |
|
Hammer Bro. posted:The results look like rear end and the first test run took an hour and a half but I typed in words and I got a 3D model! How many tris/polygons are in that burger? Could you link the model file itself so I can see what it's made of?
|
# ? Dec 2, 2022 22:42 |
|
Mercury_Storm posted:How many tris/polygons are in that burger? Could you link the model file itself so I can see what it's made of? you can see examples here: https://github.com/ashawkey/stable-dreamfusion/issues/1 based on those I don't think it's creating models in the traditional polygonal way, it's doing something weird like a bunch of intersecting planes with alpha textures
|
# ? Dec 2, 2022 22:45 |
|
|
# ? Dec 2, 2022 23:26 |
|
mcbexx posted:I am using AI images to teach myself photo retouching. The fact that so many details are hosed up is great for giving you little things to practice on. Fixing up eyes, teeth, fingers and removing the occasional extra limb are great exercises. If you don't care about it being online, Photopea is basically Photoshop but free.
|
# ? Dec 2, 2022 23:53 |
|
i can put my arm back on, but you can't. so play safe
|
# ? Dec 2, 2022 23:55 |
|
Gromit posted:If you don't care about it being online, Photopea is basically Photoshop but free. I tried to switch from Photoshop to gimp and was constantly frustrated by the workflows and missing features (even after installing a bunch of extra poo poo to make it more like photoshop) I ignored photopea for a long time because a webapp knock off of photoshop sounds like the worst but it's GREAT I've been using it for months now and I love it
|
# ? Dec 2, 2022 23:59 |
|
TIP posted:I tried to switch from Photoshop to gimp and was constantly frustrated by the workflows and missing features (even after installing a bunch of extra poo poo to make it more like photoshop) GIMP is terrible. Krita is actually an alright photoshop replacement though, and it has a stable diffusion plugin.
|
# ? Dec 3, 2022 00:25 |
|
Photoshop Elements 2022 (and Premier) is on sale on woot.com today for 67 bucks as well. Disc version, no Cloud BS.
|
# ? Dec 3, 2022 00:35 |
|
GIMP is always on sale
|
# ? Dec 3, 2022 00:36 |
|
Support for the new .safetensors format landed in webui a bit ago. Now you can download all the shady model weights you like without worrying about potential viruses or code-execution exploits since unlike a .ckpt file it only contains weight numbers and not executable code. Plus it loads significantly faster than .ckpt files. For example there is now a .safetensors download for the Anything-v3 model that a few people were suspicious of, IIRC its .ckpt version set off some false positives on some antivirus. RPATDO_LAMD fucked around with this message at 00:54 on Dec 3, 2022 |
# ? Dec 3, 2022 00:36 |
|
Mozi posted:Photoshop Elements 2022 (and Premier) is on sale on woot.com today for 67 bucks as well. Disc version, no Cloud BS. I was curious what's missing from Photoshop elements and... ouch: Features: Many of the Photoshop features are missing including the following: – Lack of 16-bit support in Photoshop Elements which results in loss in functionality of the Magic Wand, Selection Brush, the family of Type tools, Recompose, Cookie Cutter, Straighten, the brush based tools (Paint Brush, Eraser, Dodge, Blur, etc. and the shape based tools (Rectangle, Ellipse, Custom Shape, etc.). Many filters do not work either. – Smart Objects are not fully supported which means you cannot create or open embedded files as Smart Objects – There are limitations with the Color Picker functions – There is no support for a Channels panel or a Channel Mixer Adjustment/Enhancement – Many vector based tools are not available including the Pen family of tools and there is no Paths panel.
|
# ? Dec 3, 2022 00:39 |
|
TIP posted:I was curious what's missing from Photoshop elements and... ouch: Apparently the 2022 Photoshop Elements has fewer features than the 2012 Photoshop CS6, which was the last 'standalone' version of photoshop they released before switching to the creative cloud subscription-only model.
|
# ? Dec 3, 2022 00:44 |
|
RPATDO_LAMD posted:Support for the new .safetensors format landed in webui a bit ago. Now you can download all the shady model weights you like without worrying about potential viruses or code-execution exploits since unlike a .ckpt file it only contains weight numbers and not executable code. also, anything V3 is the one i'm getting these nutty looking quality results from pro tip for it to get that style (realistic, sketch), masterpiece, best quality, (subsurface scattering skin pores:1.33), (hair shader:0.3), <<YOUR IMPORTANT STUFF HERE>> realistic proportions, highly detailed, smooth, sharp focus, 8k, ray tracing, digital painting, concept art illustration by artgerm and greg rutkowski and alphonse mucha, by guweiz and wlop and artgerm, by Akihiko Yoshida, Mark Arian, Sakimi Chan, Mogoon, (concept art:1.2), <<ADDITIONAL BITS LIKE BACKGROUNDS AND STUFF HERE>>
|
# ? Dec 3, 2022 00:48 |
|
RPATDO_LAMD posted:Apparently the 2022 Photoshop Elements has fewer features than the 2012 Photoshop CS6, which was the last 'standalone' version of photoshop they released before switching to the creative cloud subscription-only model. Oh, yikes. Good to know!
|
# ? Dec 3, 2022 00:49 |
|
RPATDO_LAMD posted:Apparently the 2022 Photoshop Elements has fewer features than the 2012 Photoshop CS6, which was the last 'standalone' version of photoshop they released before switching to the creative cloud subscription-only model. "installing photoshop CS6 2016"
|
# ? Dec 3, 2022 00:55 |
|
|
# ? May 22, 2024 17:00 |
|
Does it count as abandonware if you can't legally purchase it anymore?
|
# ? Dec 3, 2022 00:57 |