|
production still from 1976 of Alejandro Jodorowsky’s Spaceballs (via reddit user dag)
|
# ? Jan 16, 2023 11:23 |
|
|
# ? May 30, 2024 01:04 |
|
I think I pretty much nailed it, I'm gonna make a few more models with some tweaks but I think I got a depth model trained on my friends face that's going to work for replacing people in videos with him and making embarrassing videos where he's a Japanese pop star. Way, way better results than my first model. It just seems like it takes a lot more training to get a depth model to work right. This is from a 4000 step 400 class image 50 training image model. I have a 250 training image set of him too but I think with more selection in the pictures it comes out better. BrainDance fucked around with this message at 12:01 on Jan 16, 2023 |
# ? Jan 16, 2023 11:59 |
|
Voidships of the 15th Imperium
|
# ? Jan 16, 2023 16:41 |
|
Is that SD or midjourney?
|
# ? Jan 16, 2023 19:21 |
|
Rutibex posted:stuff I've been generating screenshots of fake media. write an in-depth plot description of a sequel to The Thing (1982) where The Thing joins a japanese girl idol group while trying to evade the millitary posted:The Thing, now free from the ice and seeking its own freedom, finds itself approached by The Idol Star Agency, a Japanese pop music girl idol group seeking a new “edgy” member. Yes, you read that right: a girl idol group. Initially apprehensive, the Thing soon falls in love with the idea of a clean life and perfect identity, permanently in disguise as its new “Ai Nonaka” persona. write a in-description of reboot of Dexter's Laboratory as a teen drama on the CW network posted:The CW reimagines the beloved Nickelodeon cartoon Dexter's Lab in an edgy teen drama! Dexter is a 15-year-old science genius navigating the complicated social hierarchy of high school and his ever-changing home life. No longer a bumbling lab rat, this teenage Dexter is a typically charismatic, boundary pushing genius who uses his invention to fix the messes of his own wrong-doings and those around him. With a curious combination of smarts, charm, and vulnerability, he races to solve any problem that gets in his way with the help of his fierce older sister Dee Dee and his eccentric grandmother, the eccentric Professor Williams. With their power base growing, they frequently come into conflict with the rich and powerful Astronomanov family and incur the wrath of the school’s principal, Ms. Keane. The show promises to keep us on the edge of our seats as we follow Dexter and his team as they solve the science themes while wrestling with teenage angst, social power play, family ups and downs, and hormones. Dee Dee Mandark Misc characters I found by looking at the wiki Koosalagoopagoop Laboratory Jack Nicholson starring in At The Mountains of Madness
|
# ? Jan 16, 2023 20:10 |
|
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 |
|
Random prompts and some were worked on for a few img2img iterations while looking for avatar ideas. I don't know why epic diffusion kept giving me spider people for spider but I went with it. I tried to take that last one someplace but decided while an interesting piece I ultimately wasn't going to use it as an avatar even if completed. I think I want to take a step back to less humanoid.
|
# ? Jan 16, 2023 21:44 |
|
"Gaston, eat the eggs"quote:In the animated children's movie, there is a manly man named Gaston who is known for his strength and bravery. One day, while out on a walk, Gaston comes across a passing witch. Without thinking, Gaston insults the witch, calling her old and ugly. The witch, angered by Gaston's words, curses him with a spell. "Gaston, sing with the eggs" quote:Verse 1: "Gaston, rejected alternate ending" quote:As the girl gets to know Gaston, she starts to see the good within him, despite his eggshell exterior. She falls in love with him, and Gaston starts to feel the same way. But he is still a giant egg beast, and he is worried that the girl will never accept him as he is. The Butcher fucked around with this message at 23:19 on Jan 16, 2023 |
# ? Jan 16, 2023 22:56 |
|
Is there a particular preferred way to install and run Stable Diffusion et al locally? I now have a GPU with enough VRAM to run one of these models
|
# ? Jan 16, 2023 23:42 |
|
The most confusing parts of this class-action complaint against SD/MJ/etc are when they go out of their way to poo poo on the *quality* of the outputs in a way that seriously undermines their main case Implicitly gesturing at an output-based standard like this would be leading with your chin even if Midjourney v4 wasn't ALREADY trivially able to handle this prompt. I can generate hundreds of these, way more than there are photos of dogs in caps eating ice cream. Don't put stuff in a complaint just because it makes your clients fist-pump.
|
# ? Jan 16, 2023 23:42 |
|
Elotana posted:The most confusing parts of this class-action complaint against SD/MJ/etc are when they go out of their way to poo poo on the *quality* of the outputs in a way that seriously undermines their main case There's a weird edge of pettiness right from the start with the inclusion of DeviantArt as part of the defendants. Like, instead of going after OpenAI and Dall-E for a third example, they specifically chose a platform they have a personal grudge with and just happened to have declared their intent for their own AI image generation project. Presumably because ArtStation/Epic haven't actually done that last part yet. If this actually goes to trial, it feels like it's going to come down entirely to how much each side is able to persuade the judge(s) of their technical definition of what this technology actually does under the hood. Assuming they don't just settle out of court, though it seems if they did that the whole Art Twitter contingent would view these guys as class traitors and blacklist them or something. Probably the best outcome anyone can realistically hope for is the case is just dismissed without trial.
|
# ? Jan 17, 2023 00:26 |
|
Without even one concrete example of an infringing (i.e., overfit) output, it seems like at least half of these causes ought fail to clear a 12(b)(6) motion. You can call things "derivative works" until you're blue in the face, but without the words "substantially similar," I don't know where that gets you.
|
# ? Jan 17, 2023 00:36 |
|
Sir Mat of Dickie posted:Is there a particular preferred way to install and run Stable Diffusion et al locally? I now have a GPU with enough VRAM to run one of these models There's auto1111, which is being developed by a crypto-racist eastern european weirdo, but has the most features by far. It's also a pain to install, but it's still the most popular UI just because of how popular it is. There's cmdr2 which is a nice little meat-and-potatoes UI which I personally prefer most of the time, when I don't need a specific feature from Auto1111. Should be pretty easy to install. Finally, InvokeAI. Very pretty and has a very nice inpainting/outpainting UI, but somewhat wonky under the hood at the moment. I've had a lot of technical issues getting it to work, but the outpainting is, indeed, very cool. I have all three on my drives and use all three, depending on what I want to do.
|
# ? Jan 17, 2023 00:59 |
|
Megazver posted:There's auto1111, which is being developed by a crypto-racist eastern european weirdo, but has the most features by far. It's also a pain to install, but it's still the most popular UI just because of how popular it is. Thanks for the links. I saw Auto1111's repo, I wanted to know if people had reasons to avoid it besides anything about the author, but if it actually works I might use it. Edit: LOL, I installed the latest drivers and have CUDA 12.0, which means I can't install or build any of the ML frameworks because CuDNN hasn't come out for CUDA 12.0 yet. I don't want to mess with drivers and downgrade, so I'll just wait for that. Sir Mat of Dickie fucked around with this message at 04:34 on Jan 17, 2023 |
# ? Jan 17, 2023 04:20 |
|
Anyone get dreambooth ckpt training working on an M1 macbook? I'm going through tutorials for installing tensorflow locally since it seems to be missing from pip, but I don't know how to access the library from a different repo (automatic1111)
|
# ? Jan 17, 2023 05:23 |
|
Batman and The Joker in a loving, mutually beneficial relationship. Suburban bliss. Norman Rockwell.
|
# ? Jan 17, 2023 06:34 |
|
Alris posted:Batman and The Joker in a loving, mutually beneficial relationship. Suburban bliss. Norman Rockwell. This one is excellent, just need to give Batman a black glove and fix the wonky finger.
|
# ? Jan 17, 2023 07:08 |
|
Rutibex posted:Voidships of the 15th Imperium Whoa, this one is incredible. Did you explicitly tell it to strap a mech to the side of the ship or make a scavenger-themed picture or did it just come up with that itself?
|
# ? Jan 17, 2023 07:16 |
|
Some of those are clearly Robin cosplaying as The Joker
|
# ? Jan 17, 2023 07:19 |
|
Analytic Engine posted:Anyone get dreambooth ckpt training working on an M1 macbook? I'm going through tutorials for installing tensorflow locally since it seems to be missing from pip, but I don't know how to access the library from a different repo (automatic1111) In theory you could activate the venv and install tensorflow-macos and tensorflow-metal but that doesn't really do anything Ruffian Price fucked around with this message at 07:54 on Jan 17, 2023 |
# ? Jan 17, 2023 07:20 |
|
MuffiTuffiWuffi posted:Whoa, this one is incredible. Did you explicitly tell it to strap a mech to the side of the ship or make a scavenger-themed picture or did it just come up with that itself? I was just asking for spaceships, midjourney is the creative one here
|
# ? Jan 17, 2023 08:16 |
|
This stuff is insanely addictive Alex Grey's concept art for a modern Hollywood remake of Akira Kurosawa's Seven Samurai
|
# ? Jan 17, 2023 08:37 |
|
Skinner box escapism was a good way to put it. Maybe there's something even cooler just a seed away
|
# ? Jan 17, 2023 09:17 |
|
Sir Mat of Dickie posted:Thanks for the links. I saw Auto1111's repo, I wanted to know if people had reasons to avoid it besides anything about the author, but if it actually works I might use it. You can try Camenduru's Offline Installer, afaik it sets up an independent environment including CUDA. Idk what I'm doing and idk how it worked for me when I hadn't installed CUDA, but it worked*. *Not as described (run.bat just gave me an error which I couldn't use for troubleshooting) but webui-user.bat ran just fine. Ruffian Price posted:Skinner box escapism was a good way to put it. Maybe there's something even cooler just a seed away Prompt: trapped inside an AI-Generated Skinner Box endlessly producing uncanny pictures
|
# ? Jan 17, 2023 10:33 |
|
The Sausages posted:You can try Camenduru's Offline Installer, afaik it sets up an independent environment including CUDA. Idk what I'm doing and idk how it worked for me when I hadn't installed CUDA, but it worked*. I am pretty sure automatic1111 does something similar, though I havent checked, because when I was messing with diffusers and tweaking things to get it to work I had uninstalled CUDA multiple times, switched versions multiple times, and it broke near everything, torch.cuda.is_available returned false and such, yet through it all automatic1111 kept running just fine. Every single AI project is a dependency rats nest now though, all held together by gum and paper clips one upgrade away from the whole thing collapsing. ShivamShrirao's diffusers is currently, apparently, semi-broken and he hasn't been around for a few days to fix it. His dreambooth still trains, it's just off. As far as anyone can tell it's all because this was actually not a good idea and by going to accelerate 0.15.0 things are broken in a hard to pin down way, but changing a single line back and downgrading back to accelerate 0.12.0 fixes it. Older versions of cuda don't install in some distros with the version of gcc they ship with, which leads to some solutions that arent super clean. Stuff like that, it's all just kinda unstable. BrainDance fucked around with this message at 11:54 on Jan 17, 2023 |
# ? Jan 17, 2023 11:46 |
|
Sir Mat of Dickie posted:Edit: LOL, I installed the latest drivers and have CUDA 12.0, which means I can't install or build any of the ML frameworks because CuDNN hasn't come out for CUDA 12.0 yet. I don't want to mess with drivers and downgrade, so I'll just wait for that. The other two installers just install their own environments by default, instead of being a piece of poo poo like Auto1111. Have you actually tried the other two?
|
# ? Jan 17, 2023 11:54 |
|
Ruffian Price posted:The dreambooth extension got updated for M1 in November, paste the github in the Extensions tab and you should be golden, ignore the tensorflow errors and train on CPU. 2700 steps took me nine hours on M1 Max and it was overtrained anyway, leave your laptop overnight (don't forget to turn on Amphetamine) with the "save checkpoint every 500 steps" option on and then compare results Thanks! I used the URL and installed it, but the app UI is missing the dreambooth tab after restarting. The tensorflow error appears in the zsh logs at startup and seems to be why the extension is hidden
|
# ? Jan 17, 2023 12:34 |
|
RPATDO_LAMD posted:the whole page is here: https://stablediffusionlitigation.com/ An here's a response to the points they make in the complaint: http://www.stablediffusionfrivolous.com/
|
# ? Jan 17, 2023 13:08 |
|
https://twitter.com/arawnsley/status/1615140587010916353?s=20&t=ZunhNSIIplnl5AY4MbuS5A
|
# ? Jan 17, 2023 13:36 |
|
Elotana posted:Without even one concrete example of an infringing (i.e., overfit) output, it seems like at least half of these causes ought fail to clear a 12(b)(6) motion. You can call things "derivative works" until you're blue in the face, but without the words "substantially similar," I don't know where that gets you. I don't get it how is their shining example not mona lisa or that it steals the style of Van Gogh. Like you could probably convince a jury that something there is wrong. But the examples given? These people know what they don't like about the tool and it's not the copying. They don't like the lack of human in the creation. It's quite simply technophobia.
|
# ? Jan 17, 2023 14:16 |
|
The lawsuit is pointless. it doesn't even matter if you force them to make a squeaky clean image AI who has only ever seen classical paintings. all it takes is one person to retrain it on a few pics of Greg Rutkowski Unless they make possessing a transformer models as illegal as hard drugs they cant stop people from retraining it. Heck they can't stop it even if they go that far now that stable diffusion exists, only push it underground where it will be used exclusively for art crime. Disney will have their own in-house AI and they will licence it out and artists will still have no work
|
# ? Jan 17, 2023 14:26 |
|
or they'll just be hosted in japan where copyright law has already added an explicit fair-use carveout for training computer models
|
# ? Jan 17, 2023 14:30 |
|
Tunicate posted:or they'll just be hosted in japan where copyright law has already added an explicit fair-use carveout for training computer models That's actually really surprising because otherwise Japanese copyright law is insanely draconian
|
# ? Jan 17, 2023 14:46 |
|
Honestly, this lawsuit is probably going to do more good for AI than bad if it goes to trial. I don't see how they actually can win it, especially with the argument they are putting forth. Heck, Sarah Anderson sabotaged the entire thing when she put out that article that admits AI can't actually reproduce her art, just vaguely look like her style (which is not actually copyrightable). Also, if they try to bring up Greg Rutkowski, the training dataset actually didn't contain any of his art in the first place. It was just art being classified to be by him since it looked similar to his style. All this will do is set a legal precedent of fair use. Even if the judge is some eldritch being who cringes at the mere thought of being forced to use a computer, they aren't going to be able to come up with specific art that infringes. Heck, an old judge might even work against them because process is not copyrightable, so they aren't really going to care how the technology works, but the results.
|
# ? Jan 17, 2023 14:58 |
|
BrainDance posted:That's actually really surprising because otherwise Japanese copyright law is insanely draconian everyone in Japan is a pervert. no one is going to ban the magical hentai machine
|
# ? Jan 17, 2023 15:37 |
|
|
# ? Jan 17, 2023 16:19 |
|
pixaal posted:I don't get it how is their shining example not mona lisa or that it steals the style of Van Gogh. Like you could probably convince a jury that something there is wrong. For the former, da Vinci died five centuries ago, and the latter, dude died 133 years ago; how could any jury find legal wrongdoing there? "Style" is not and has never been copyrightable
|
# ? Jan 17, 2023 16:39 |
|
Fuschia tude posted:For the former, da Vinci died five centuries ago, and the latter, dude died 133 years ago; how could any jury find legal wrongdoing there? "Style" is not and has never been copyrightable 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.
|
# ? Jan 17, 2023 16:42 |
|
I am not getting good results on the SD discord, oh well: prompt:gamer guy, computer desk, wearing fedora, mountain dew bottles, artstyle : patrick nigel
|
# ? Jan 17, 2023 16:45 |
|
|
# ? May 30, 2024 01:04 |
|
"AI steals real artists' work!" Me, making AI images:
|
# ? Jan 17, 2023 16:58 |