|
BARONS CYBER SKULL posted:close up painting, vampire, sci-fi armor, purple smoke hair, beautiful, stern face, gorgeous, face visible, scars, eyeliner, cinematic lighting, eva suit, wallpaper, extremely detailed, sharp focus, by Greg Rutkowski and [[[WLOP]]], intricate, beautiful, award winning, Trending on artstation, pixiv Weird how SD assumes the subject's gender. Problematic. Adding "male" to your prompt.   
|
#
¿
Aug 31, 2022 00:12
|
|
|
|
| # ¿ May 14, 2024 14:33 |
|
|
Bottom Liner posted:the more pressing issue is that literally all of these default to white unless you use certain language that is uh, not great. Eeeeh, the last one could be interpreted as being asian at least, but yeah, you're not wrong. But phew, at least it is not actively ignoring color of skin when properly prompted, so ... yaaaaay? And it looks the coolest, so maybe the AI is trying to make up for its inherent caucasian bias. As someone who can't draw/paint at all, this poo poo is blowing my mind. No one is ever going to hire someone to do an album cover ever again, are they? 
mcbexx fucked around with this message at 00:42 on Aug 31, 2022 |
|
#
¿
Aug 31, 2022 00:27
|
|
|
Any idea why the local SD installation sometimes overwrites the existing output file when you go back and do an iteration with slight differences? Sometimes it does, sometimes it doesn't and I can't figure out what triggers it. The output files are numbered, 00000.png counting up and sometimes it just hops back a couple of digits and starts overwriting. mcbexx fucked around with this message at 01:00 on Aug 31, 2022 |
|
#
¿
Aug 31, 2022 00:56
|
|
|
Can we, as a self-imposed, unenforced rule, try to add some basic info to the image posts? Just the name (or a short tag) for the engine and the prompt that were used? I think that would be helpful to figure out how certain keywords affect the output.
|
|
#
¿
Sep 1, 2022 03:21
|
|
|
Bula Vinaka posted:https://www.youtube.com/watch?v=L9hlfc16qg0&t=565s This is a new medium. It may not be other artists' medium of choice, but it is a tool to turn ideas into images and, yes, ultimately, art. Everyone can take a paint brush and splash paint on a canvas like Pollock did - or in a way that resembles his work. Why is his work considered art and mine most likely wouldn't be? Did he control every single random splash to get the exact outcome he intended? Where do you draw the line between randomness and intent when it comes to art? Is reproducibility an issue? Warhol used screen printing to mass produce his art. It was part of his process.
|
|
#
¿
Sep 2, 2022 22:45
|
|
|
Lord Stimperor posted:Also turns out that Stable Diffusion is absolutely not opposed to drawing, uh, tasteful act portraits for, uh, scholarly pursuits It also does not shy away from adding the occasional surprise third boob (clothed or otherwise), completely unprompted.
|
|
#
¿
Sep 3, 2022 23:29
|
|
|
Found this excellent SD prompt and messed around with it, I think its output is insanely good. Replace subject and add attributes like muscular, athletic, obese as needed. Obese (with everything else unchanged) gave me a fully nude output though. Weird. detailed photo of an old bronze patina statue of beautiful lara croft, full body portrait, photorealism, intricate detail, museum diffuse lighting -C 10 -H768    Throwing in Jackman's Wolverine for good measure:  Seems like SD is really good at anatomy. Well, most of the time. (Do I need to NSFW this?) 
|
|
#
¿
Sep 4, 2022 02:26
|
|
|
TheWorldsaStage posted:I enjoy that SD knew Logan needed daisy dukes Then you'll probably love this:   I noticed that with a lot of characters (this one, Deadpool, for instance), their color scheme completely overrides the "bronze patina" parameter. How can I put emphasis on that particular part? I tried adding opening/closing parentheses, but that didn't do anything. All Hulks came out green, and not bronze patina green.
|
|
#
¿
Sep 4, 2022 12:11
|
|
|
Here's a handy guide on artists' styles for SD. Thanks, concerned artist! https://twitter.com/arvalis/status/1558632898336501761
|
|
#
¿
Sep 7, 2022 15:02
|
|
|
quote:*Celebrity name*, as hulk, digital painting, artstation, concept art, smooth, sharp focus, illustration, art by artgerm and donato giancola and Joseph Christian Leyendecker, Ross Tran, WLOP -H704 -C 10 is a source of endless joy. Caveat warning: The output for female celebrities gets quite horny pretty quick. Complete Gallery with highlights - possibly - here. I'll probably keep updating the gallery until I get bored.Edit: Got a bit carried away, 2 hours later its 50+ images and counting  You wouldn't like them when they're angry.     Tatiana Maslany lol  And finally, just to make sure I'm going to hell. 
mcbexx fucked around with this message at 12:24 on Sep 14, 2022 |
|
#
¿
Sep 14, 2022 01:40
|
|
|
Day 3, still not tired of creating celebrities hulking out. I added about new 50 images since my last post. Updated gallery here, including: Danny Trejo  Robert de Niro  and guest starring, Bruce Willis as Thanos 
|
|
#
¿
Sep 16, 2022 16:34
|
|
|
I have been using the Lstein repo and I have hardly any trouble with extra faces when generating 512x704 images and using the built in upscaler - both 2x and 4x look very clean (2070S 8GB). It has a web UI as well, but I hardly ever bother, I'm mostly feeding text files with a list of prompts and variations to the CLI.
|
|
#
¿
Sep 23, 2022 03:08
|
|
|
Am I understanding this correctly, k_euler(_a) is the only of the current sampling methods which will not be able to reproduce a previous prompt when providing the exact seed number?
|
|
#
¿
Sep 25, 2022 15:45
|
|
|
Rutibex posted:"pinup girl, large feet, made of yogurt" You had to go there, didn't you? 
|
|
#
¿
Sep 27, 2022 16:16
|
|
|
Found a limited time offer for canvas prints, 35" x 23.6" for $22. Helloooo Etsy! Or everyone I know will get "unique AI art" for the next 5 christmases. Really tempted to see how this would look printed and if it manages to keep the vibrant colors. Edit: Oh poo poo, I only just noticed, probably not exactly this one.  ("Cyberpunk Red Light District")  "PICE" "PID ULT" 
mcbexx fucked around with this message at 18:54 on Sep 27, 2022 |
|
#
¿
Sep 27, 2022 17:46
|
|
|
WhiteHowler posted:And I haven't even gone into GIMP yet to try to fix the character's eye and smooth out the masking around her face. IMHO that's also part of the fun (and "creative process", if you will) - fixing up stuff, "frankensteining" parts from different variations into a satisfying composition etc. I know I am getting a lot of practice and experience from futzing with my output. I feel "artists" who huff and puff over it don't really understand what it has to offer, especially to people who are not gifted with the manual skillset, but now have a tool to express their imagination.
|
|
#
¿
Sep 27, 2022 18:35
|
|
|
"LICFCKE" "Party's over, the po-po is here." 
mcbexx fucked around with this message at 20:11 on Sep 27, 2022 |
|
#
¿
Sep 27, 2022 19:46
|
|
|
My main dreamlog.txt file, which stores all prompts along with the corresponding seed# and output filename is just a couple of bytes shy of 3 MB and contains 7931 lines. It has been created 11 days ago.
|
|
#
¿
Sep 27, 2022 21:28
|
|
|
Training your own models in the cloud for as little as $0.30 per hour (that's about one prompt "term", depending on complexity). Get yourself, your mom, dad, girlfriend/boyfriend, dog, archnemesis into your images. Mind: boggled. https://www.youtube.com/watch?v=7m__xadX0z0
|
|
#
¿
Sep 28, 2022 22:10
|
|
|
I hope someone can find a way around the problem with extra anatomy on images with certain aspect ratios or exceeding the resolution threshold by too much. Sucks that most repos now can handle 1024x768 with 8 GB of VRAM, but even invoking negative prompts to avoid deformities and additional body parts just stacks on extra heads 9 out of 10 times or creates abominations that would make Cronenberg break down in horror. Bummer. Exceeding 576x768 seems to be where it'll start to get weird quite more often. Or are there any workarounds for this?
|
|
#
¿
Oct 12, 2022 16:50
|
|
|
(removed)
mcbexx fucked around with this message at 21:41 on Apr 19, 2023 |
|
#
¿
Oct 16, 2022 03:48
|
|
|
Gotta say, not a big fan of the 1.5 model. I often use high cfg values (16 usually) and 100 steps for realistic and detailed facial features/textures and the outcomes across the board are way less convincing/aesthetically pleasing than with 1.4 (same prompts and seeds). Meh. Backup all your model files. I just hope all the different clients will keep backward compatibility going forward.
|
|
#
¿
Oct 27, 2022 00:21
|
|
|
I'm using the InvokeAI WebUI and they posted an announcement video for their upcoming 2.2 release. I'm not following other WebUIs that closely and the InvokeAI guys are in general a bit slower with their release/feature cycles, so maybe this is old news and boring, but this giant canvas thing looks neat. https://www.youtube.com/watch?v=hIYBfDtKaus
|
|
#
¿
Dec 1, 2022 02:54
|
|
|
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
|
|
|
Mozi posted:Photoshop Elements 2022 (and Premier) is on sale on woot.com today for 67 bucks as well. Disc version, no Cloud BS. Again, before anyone spends that much on a feature limited lite version of PS, take the free trial version of Affinity Photo for a spin. I may sound like a shill, but it's really powerful and comprehensive, with tons of tutorials available on YT. It's been on sale for 50% off lots of times from the developers in the past (I paid 24,95 for my v1.x licence back then).
|
|
#
¿
Dec 3, 2022 05:01
|
|
|
Kharmakazy posted:CHOOSE YOUR JOHN HENRY TO FIGHT THE STEAM POWERED AI ART MACHINE So we meet again, Judohobo. It has been a while. 
|
|
#
¿
Dec 5, 2022 02:45
|
|
|
Rutibex posted:You can also find various kind of stable diffusion around, and its quite nice too. But thats high nerd stuff with like special java versions and all kind of annoying things The more popular installations of Stable Diffusion, namely Automatic1111 and InvokeAI come with one-click install scripts that do all the annoying nerdy things for you. Sure, if for some obscure reason something breaks on your very specific computer config you're hosed if you're not technically inclined (or can't muster the patience to read through a page of troubleshooting advice or search for a step-by-step video tutorial on YT outlining possible pitfalls), but setting up SD@home is (or can be) a very painless endeavor.
|
|
#
¿
Dec 6, 2022 12:39
|
|
|
NSFW that butthole pls. Edit: was phoneposting and edited in the wrong image(s), my bad. mcbexx fucked around with this message at 16:09 on Dec 8, 2022 |
|
#
¿
Dec 8, 2022 16:05
|
|
|
Someone made an add-on for Blender that uses SD to texture 3D-models (whole scenes, actually). Pretty cool for LOD models/textures or prototyping. https://i.imgur.com/xkQW2fY.mp4 Link to reddit post with more information from the OP. mcbexx fucked around with this message at 23:39 on Dec 15, 2022 |
|
#
¿
Dec 15, 2022 23:37
|
|
|
Invoke AI has put up a new timelapse video featuring their unified canvas, it's a fun watch and gives a good impression on how it manages to set up large and consistent images. https://www.youtube.com/watch?v=WmOUl8Gab5U Oh, there's an actual part 1 as well, I thought this was a continuation of their release video. Edit: Quick image made with the 2.5 canva, just testing some of the inpainting/outpainting features and fiddling with the seam correction sliders to get a feel for how it works. 
mcbexx fucked around with this message at 03:41 on Jan 2, 2023 |
|
#
¿
Jan 2, 2023 02:10
|
|
|
This is like a Bob Ross matinee for me. I love watching this poo poo. They also snuck in an announcement for some big news coming for InvokeAI in the end. Probably SD 2.x support. Or their Nodes based workflow, which has had its placeholder in the WebUI since v2.5(?). Either way, check out InvokeAI, y'all. https://www.youtube.com/watch?v=e_rRQeee6-0
|
|
#
¿
Jan 15, 2023 06:42
|
|
|
The more I'm playing around with masking and inpainting, the more I'm jonesing for a 4090. Especially since for the final touches, I found that upscaling an image and then using the result in the unified canvas to in-paint a better quality high-resolution face in. Much better than rendering something first at a feasible resolution and just running GPFGAN and upscaling. Means that the bounding box/mask for a face can easily be 1000x1000 pixels, which is now taking so long on a 2070S/8GB that it's starting to bother me.
|
|
#
¿
Jan 16, 2023 20:57
|
|
|
pixaal posted:Oh I'm not calling them slam dunks I'm saying they'd have at least some chance of not being laughed out of the court. Both of your examples are public domain though? They would not even be laughed out of court, they would not even be enter to go near one if they based their claims on that Not a lawyer, feel free to correct me.
|
|
#
¿
Jan 17, 2023 17:04
|
|
|
deep dish peat moss posted:"AI steals real artists' work!" Am I the only one finding it ironic that many "real artists" who - sometimes exclusively - publish fanart of widely popular intellectual properties (admittedly pretty great art, from an artisinal standpoint, don't get me wrong), are the most vocal about "AI being theft"? I would wager that within a year, the majority of (digital) artists will be using AI to create concepts/assets they will manually refine , without batting an eye or even remembering that they huffed and puffed about it today. This one caught my eye and gave me a good chuckle. "Beep boop, no humans are involved in AI generated art, just us computers, beep boop!" 
|
|
#
¿
Jan 17, 2023 22:32
|
|
|
Ooh, I got some movie stills as well, from a never released 70's prequel to "The Shining".  
|
|
#
¿
Jan 19, 2023 23:32
|
|
|
AARD VARKMAN posted:It's like an entirely fun new sandbox with how intuitive it is, I can't wait for full photoshop-level capabilities to eventually be available with this, there's a lot of little stuff I'd like to do with basic image manipulation between generations. But even the masking + outpainting/inpainting combo is a blast. In case you haven't tried, unified canvas fully works with the clipboard. So nothing is stopping you from right-clicking and copying your canvas image at full size, pasting it into Photoshop/Affinity Photo/whatever, do some fine tuning or collaging (especially useful for img2img), copy flattened and paste it back into the canvas. Speeds up things nicely. If you want to make a snapshot, just save the canvas or just your bounding box to your gallery.
|
|
#
¿
Jan 24, 2023 04:42
|
|
|
Someone asked me if I could make them a character portrait for their upcoming D&D campaign and it still amazes me that with three short lines of text (and a style and negative prompt that I basically keep on file) I can get a dozen variations of this in under 5 minutes. They were so happy with the result and it took next to no effort. I like making people happy.   Even the unintended/unprompted outcomes were amazing in their own way.  
|
|
#
¿
Jan 25, 2023 18:15
|
|
|
Some cyberpunk-y characters. Including "movie poster" in the prompt did a lot of heavy lifting for the second one.  
|
|
#
¿
Jan 26, 2023 16:19
|
|
|
Doctor Zero posted:Dang. What did you use for that? InvokeAI + Protogen_X 3.4 checkpoint model and a second pass with just using inpainting on the faces after 4x upscaling the original 576x768 image. I found that helps way better with face wonkyness than face restoration and adds a lot of detail and skin texture (which upscaling usually smoothes out way too much) Megazver posted:Looks like one of the Protogen models. Yep. I found that once you use that, it's really rough to go back to SD 1.x.
|
|
#
¿
Jan 26, 2023 17:49
|
|
|
|
| # ¿ May 14, 2024 14:33 |
|
|
Cabbages and Kings posted:do you just mean vanilla 1.x? Isn't Protogen itself built off 1.5, not 2.x? Yes, plain 1.4/1.5.
|
|
#
¿
Jan 28, 2023 15:56
|
|