|
smoobles posted:I was trying to generate Heathcliff comics and Bing gave me this instead. LOL. At least it's literary. axolotl farmer posted:Bing's unsafe warnings are all over the place. How much you want to bet it's using an LLM (probably ChatGPT) to detect unsafe content. I would expect that would be unreliable at best, just like that. Tarkus posted:Does anyone else feel almost dirty generating these images? I'm not even talking about art theft because I'm not sure that it really is theft. I almost feel like as though I am in art school in, say 1850, and I come to class one day with a digital camera and a printer. I find the images it generates are usually visually appealing but I feel nothing for them. Weird. Welcome to the AI revolution. FINALLY In the Battle of AI art, ChatGPT with Dall-e 3 is the WINNER   No other service has been able to give me this. I am completely satisfied. 
|
#
?
Oct 12, 2023 18:16
|
|
|
|
| # ? May 28, 2024 00:49 |
|
|
Doctor Zero posted:LOL. At least it's literary. Why does he have 2 sets of fingers holding the mug? handle and bottom So close, but no, not perfect.
|
|
#
?
Oct 12, 2023 18:18
|
|
|
smoobles posted:I was trying to generate Heathcliff comics and Bing gave me this instead. I decided to ask GPT what is funny about these two panels. Here is what it said: quote:Alright, let's break this down. The two panels depict two contrasting scenes. On the left, we have what looks like a well-dressed individual, possibly from an earlier time period, looking calm and composed with a tea setting. The setting feels calm, perhaps a little posh, with even a cat lounging about. I was expecting it to hallucinate an answer but it did its best and found no satisfying answer. Just thought that was interesting. I suspect that the image generation GPT, when making comic panels, doesn't try to make a story or anything despite it being expected in a comic.
|
|
#
?
Oct 12, 2023 18:27
|
|
|
Tarkus posted:I decided to ask GPT what is funny about these two panels. Here is what it said: I expect "comics are supposed to convey a single story, in chronological order, with one event per panel" requires at least one level of abstraction and reasoning above where these systems are currently operating.
|
|
#
?
Oct 12, 2023 18:53
|
|
|
Fuschia tude posted:I expect "comics are supposed to convey a single story, in chronological order, with one event per panel" requires at least one level of abstraction and reasoning above where these systems are currently operating. GPT can actually do 4 panel comics! It does 1 image per panel and keeps a story that way. Dalle itself kind of makes an attempt, it gets weird when it includes 4 panel comic in the prompt and you get 4 4 panel comics. e:      that was great can we do another one? Topic of this comic Adventure Mouse explores a cave looking for treasure    
pixaal fucked around with this message at 19:20 on Oct 12, 2023 |
|
#
?
Oct 12, 2023 19:07
|
|
|
Tarkus posted:Does anyone else feel almost dirty generating these images? I'm not even talking about art theft because I'm not sure that it really is theft. I almost feel like as though I am in art school in, say 1850, and I come to class one day with a digital camera and a printer. I find the images it generates are usually visually appealing but I feel nothing for them. Weird.  When the OSHA thread just ripped rear end on a picture I suggested, it made me mad because every industry rips off copyright ESPECIALLY the construction industry. Those people are willing and do buy knockoff and counterfeit parts and tools. They reason to themselves that theyll rather buy a cheap knockoff to do one job rather than entertaining the idea of actually buying a real product. They are the reason John Deere is putting DRM on poo poo. They are the reason both online and brick retailers are filthy with counterfeits. They are the reason that sites like Alibaba exist. If you feel guilty, then don't worry because that is humanity's nature to steal and rip off.
|
|
#
?
Oct 12, 2023 19:27
|
|
|
Tarkus posted:Does anyone else feel almost dirty generating these images? I'm not even talking about art theft because I'm not sure that it really is theft. I almost feel like as though I am in art school in, say 1850, and I come to class one day with a digital camera and a printer. I find the images it generates are usually visually appealing but I feel nothing for them. Weird. It depends. Even for someone as pro-AI as myself there are lots of times I feel the output is garbage; but that is because it is. Wether it be due to imperfect models or bad prompting or just the very nature of the throw-paint-at-the-wall-and-hope-it-comes-out-right process, I find myself getting more and more annoyed when the AI gets it almost, but not quiet right. But that's the same way I feel about the photography I do. 90% of the images I take are garbage, because my hands were shaking or the apperature was 0.1 off, or I forgot to set the focus - but when I do everything correctly I get pieces I can (and have) hang/hung in galleries. With AI, if you get something 90% of the way to perfect, a quick five minutes to an hour in photoshop or what-have-you can make all the difference. For example;  "A rakshasha tiger-man monk meditating by hovering over an open flame inside a richly decorated palace chamber as the rightful king is tied up in chains" - Bing/dalle-3 Is this a cool picture? Yeah. It's better than anything I could have drawn on my own, without a serious time commitment. It's almost, but not quiet what I wanted. But it's close enough that I can fix it with a minimal amount of effort.  Now it works, and the apathy (for this image specifically) is gone. And yeah, that might be because I'm more invested in it now, but at the same time, just as with photography, every part of the process is entertaining to me, There is a joy in exploration, of finding new ways to generate images, to vary them, to compose etc. So that's one part of the dirtiness equation, finding your own fun, figuring out how to use this tool as part of your own creative process. The other part is either impostor syndrome (almost every artist thinks they are the worst artist ever) which can certainly diminish your enjoyment of the finished product regardless of its objective quality, or perhaps - as those art schoolers in the late 19th century found out - that with new methods to paint, new ways to paint must be invented to 'stand out' - since everyone and their dog could take a photograph and print it out to arbitrary size and have it perfectly resemble reality, they started going abstract. Perhaps we need to do the same, either playing to the ai's strength and going all out on dreams and impossibilities, or attempt to refine unto perfection something as simple and complex as a Normal Human Hand. tl/dr just make something fun and dumb and it'll feel better 
|
|
#
?
Oct 12, 2023 19:48
|
|
|
I'm remembering this quote from Rob Sheridan (NIN art designer and creator of the Dancing Baby meme) about when he talked to the Midjourney guys about creating horror images.
|
|
#
?
Oct 12, 2023 21:50
|
|
|
Also this is a good article. tl;dr: "So it’s not creativity, it’s not art, but it’s something else. A discovery perhaps. I discovered a corner of latent space where this image came into being. I like it, and that’s all that matters in the end, your mileage may vary." https://www.technollama.co.uk/asking-whether-ai-outputs-are-art-is-asking-the-wrong-question
|
|
#
?
Oct 12, 2023 21:52
|
|
|
I enjoy exploring latent spaces and sharing what comes out. I don't care what we call it, right now all we got is art.
|
|
#
?
Oct 12, 2023 22:41
|
|
|
Bing seems to be better today, I'd really love to know what the images that are getting filtered look like because these are some of the ones I generated:           edit. Sorry! brocked fucked around with this message at 23:33 on Oct 12, 2023 |
|
#
?
Oct 12, 2023 22:43
|
|
|
The National Gallery of Art posted:The Nineteenth Century: The Invention of Photography
|
|
#
?
Oct 12, 2023 22:48
|
|
|
https://twitter.com/MNateShyamalan/status/1712323482166722680?t=YTRQgfjM9eH9fBOZ61iv_A&s=19
|
|
#
?
Oct 12, 2023 23:04
|
|
|
|
|
#
?
Oct 12, 2023 23:17
|
|
|
brocked posted:Bing seems to be better today, I'd really love to know what the images that are getting filtered look like because these are some of the ones I generated: you're right, I retried some things that failed repeatedly yesterday and they actually worked today first try 
|
|
#
?
Oct 12, 2023 23:47
|
|
|
Shishkahuben posted:https://twitter.com/MNateShyamalan/status/1712323482166722680?t=YTRQgfjM9eH9fBOZ61iv_A&s=19 https://www.youtube.com/watch?v=H1QKyB_0P0E
|
|
#
?
Oct 12, 2023 23:55
|
|
|
gently caress yeah it is so nice that Comfy UI will pull up whatever nodes/prompts/models you used on old pics even if it was Automatic1111 images. Thought I lost some prompts forever by accidently deleting them from the log file, but you drag an image into Comfy UI and everything populates.
|
|
#
?
Oct 13, 2023 04:14
|
|
|
TIP posted:you're right, I retried some things that failed repeatedly yesterday and they actually worked today first try  Did make some decent xenomorphs in weird alien environments, but even then it would sometimes object to another run with the exact same prompt! Edit: it must have been trained on Rob Liefeld pictures, because it's doing everything possible to avoid drawing feet. Full length, full figure, long shot, whole body in frame, standing on a rug - nope, you'll get people framed from the knees up and like it! Edit edit: tried to put a xeno on the surface of Venus, then on "the second planet of the solar system" - got the 'content warning, you could be suspended' alert both times. 
Small Strange Bird fucked around with this message at 11:27 on Oct 13, 2023 |
|
#
?
Oct 13, 2023 08:01
|
|
|
yeah it may have just been luck, got a lot of lame prompt rejections later
|
|
#
?
Oct 13, 2023 08:31
|
|
|
found a weird cookbook from the 1950s and scanned the pages I present, Gregory Stipson’s Cookery Book For The Well To Do House-Woman                 
|
|
#
?
Oct 13, 2023 08:34
|
|
|
Woke Goku rules but he needs a partner... Woke Vegeta    And a foil:  
|
|
#
?
Oct 13, 2023 08:35
|
|
|
Xenos in strange alien (ahem) environments.       
|
|
#
?
Oct 13, 2023 12:19
|
|
|
I got some fun Spidermen before he got banned.     
|
|
#
?
Oct 13, 2023 14:53
|
|
|
You can probably sidestep into it, the image detection AI doesn't seem trained in it. I was attempting to get a specific person's avatar from a discord webslinging and hanging access points, but it forgot to describe their name in this image so it gave spiderman Illustration: Dynamic action scene on a skyscraper. %user% with web designs on his outfit swings using web-like ropes, installing wireless access points on different levels. Birds and helicopters capture the height.  GPT came up with this work around itself and I didn't even want real spiderman! Enjoy. I think this is the keypart "with web designs on his outfit swings using web-like ropes"
|
|
#
?
Oct 13, 2023 15:05
|
|
|
Silver-fox Spider-man
|
|
#
?
Oct 13, 2023 15:14
|
|
|
It won't let me make an image of anyone "punching a cop." What a bunch of fash bullshit. e: At least it's easy enough to get around.    
feedmyleg fucked around with this message at 15:42 on Oct 13, 2023 |
|
#
?
Oct 13, 2023 15:22
|
|
|
   
|
|
#
?
Oct 13, 2023 16:12
|
|
|
All variations on a theme.     
|
|
#
?
Oct 13, 2023 16:24
|
|
|
    e: to cut the sleaze just a little...  
feedmyleg fucked around with this message at 16:49 on Oct 13, 2023 |
|
#
?
Oct 13, 2023 16:36
|
|
|
Today I have been generating cathedrals with Bing and trying to push the impossible aesthetics. Various prompts, not all the descriptors made sense, "rose-vaulted, hyper-metallic, prismatic lighting, fractally catholic" you get the idea. I don't feel like I created these but discovering them is a lot of fun.                There's a lot of messiness to these I'd clear up if I had them in a thing I could do additional passes on them.
|
|
#
?
Oct 13, 2023 17:49
|
|
|
What the gently caress!
|
|
#
?
Oct 13, 2023 18:11
|
|
|
If you are commonly copying some things, the word "discord" is now blocked on dalle. This one took a bit to figure out (so that's GPT4 and Bing interfaces, neither know it's being blocked on the Dalle side, GPT was clueless and when I told it try not using Discord it told me that word should be safe but it would try anyway without it, it worked)
|
|
#
?
Oct 13, 2023 18:22
|
|
|
pixaal posted:If you are commonly copying some things, the word "discord" is now blocked on dalle. This one took a bit to figure out (so that's GPT4 and Bing interfaces, neither know it's being blocked on the Dalle side, GPT was clueless and when I told it try not using Discord it told me that word should be safe but it would try anyway without it, it worked) Can you get around this by asking for the Goddess Eris and the Golden Apple?
|
|
#
?
Oct 13, 2023 18:28
|
|
|
KwegiboHB posted:Can you get around this by asking for the Goddess Eris and the Golden Apple? Probably, but if you want a Discord interface it's not happening.
|
|
#
?
Oct 13, 2023 18:30
|
|
|
It really doesn't like it when you feed it specific names. Been trying to get a crowd of Abe Vigodas but to no avail.   
|
|
#
?
Oct 13, 2023 19:32
|
|
|
It's Friday the 13th in October, y'all. Time to post spooky images:   
|
|
#
?
Oct 13, 2023 19:35
|
|
|
dingo and baby eat at a table together    Have a Borkus  It does not know how to shape shift him into a bus, which is sad because old Bing could. I guess this is GPT+Dalle3, but old Bing felt a lot like something better than Dalle 2, and I'm pretty sure it was just a half trained Dalle 3, or dead end branch attempt at 3.  
pixaal fucked around with this message at 20:37 on Oct 13, 2023 |
|
#
?
Oct 13, 2023 20:08
|
|
|
If you're looking to dive into the weird and wonderful world of AI art generation using Stable Diffusion on your local PC and have a video card with at least 8GB VRAM, this weekend might be the best time to start with the newly released higher-res SDXL dataset. Here's a bunch of tips to help you on your way to setting up a basic workflow in the very flexible node-based browser-based editor ComfyUI: Download and install Stability Matrix AI suite manager that will be used for downloading custom SD models and further customizations, as well as updating ComfyUI - https://github.com/LykosAI/StabilityMatrix/releases Use the Packages tab to download the latest ComfyUI. Once finished it will be integrated into the Stability Matrix suite to allow using the Model Browser tab to download the trained datasets directly from https://civitai.com  Now onto the actual ComfyUI editor where the AI art generation will take place. First get the ComfyUI-Manager plugin: https://github.com/ltdrdata/ComfyUI-Manager/archive/refs/heads/main.zip Extract, and copy the ComfyUI-Manager-main folder into your "\StabilityMatrix\Data\Packages\ComfyUI\custom_nodes" directory. Relaunch ComfyUI and it'll download the dependencies and the Manager new button will show up in your editor.  Within the Manager click Install Custom Nodes and use the search field to install the following plugins for painless use of this noodle-based editor: (also make sure to relaunch the app so it gets the dependencies) ComfyUI Impact Pack Efficienty Nodes for ComfyUI ComfyUI-Bmad-DitryUndoRedo  From the Model Browser tab of Stability Matrix I recommend you download the following startup components: SD XL (get both the VAE Fix and the refiner VAE Fix) SDXL VAE Detail Tweaker XL realESRGAN_x4+HAT (then copy realesrganX4HAT_v10.ckpt from your \StabilityMatrix\Data\Models\StableDiffusion\ directory into the \StabilityMatrix\Data\Models\ESRGAN directory)  Once those are downloaded you should be good to go as far as having all the raw image data to manipulate and the tools to do it with offline. Now here's some basic components of the actual ComfyUI workflow that you can build upon:  The compact "all-in-one" loader is used for loading the base model, the refiner, specifying the size of the generation as well as connecting to the Low-Rank Adaptation of Large Language Model (LoRA) stack which are like contextual modifiers or "smart filters" for the generated image:  The sampler is the engine of AI generation that performs functions on the dataset to come up with a coherent image. You can play with these settings to generate wildly different images from the same promt:  FaceDetailer is an automatic inpainting module that can fix messed up faces/eyes/hands but also slightly alters the entire image quality since it does contextual blending. Hopefully a smarter one comes out soon that only blends the image around the area of the mask:  Upscaler increases the size of the image heuristically and is pretty fast, more upscalers can be downloaded from https://huggingface.co  Lastly, you can import and export workflows in ComfyUI as .json files - here's the one in the example, enjoy your journey! https://pastebin.com/raw/EHvnh8y7 Nyan Bread fucked around with this message at 21:06 on Oct 13, 2023 |
|
#
?
Oct 13, 2023 20:46
|
|
|
Bing doesn't like images of baphomet for some reason but this guy was no problemo.
|
|
#
?
Oct 13, 2023 21:00
|
|
|
|
| # ? May 28, 2024 00:49 |
|
|
Shishkahuben posted:https://twitter.com/MNateShyamalan/status/1712323482166722680?t=YTRQgfjM9eH9fBOZ61iv_A&s=19 I don't know if anyone pointed it out but that Woku on the bottom left is throwing the Entirely from a 4chan prank too. What happened to 4chan making people drill holes into their iPhones to get access to the secret 3.5mm audio jack? EVIL Gibson fucked around with this message at 21:32 on Oct 13, 2023 |
|
#
?
Oct 13, 2023 21:29
|
|