|
Have some knock-off muppets! These were stolen with Cinematic Diffusion and Analog Diffusion. Adding bisexual lighting to prompts with these models really brings out the colors. Prompts are variations on portrait photo of muppet, bisexual lighting, dim key light, exaggerated facial expression with some negative prompting to try to remove humans from the output.
|
# ? Dec 31, 2022 01:44 |
|
|
# ? May 29, 2024 23:09 |
|
Swedish Thaumocracy posted:What are some of the most striking images you lot have made since this whole thing started? I dunno, do these count as striking? quantum wavetrace of two hydrogen atoms at moment of fusion Steps: 64, Sampler: k_euler_a, CFG scale: 7.5, Seed: 3451137522, Size: 512x512, Model hash: 7460a6fa inverse quantum wavetrace of two hydrogen atoms at the moment of fusion Steps: 64, Sampler: k_euler_a, CFG scale: 7.5, Seed: 2046138787, Size: 512x512, Model hash: 7460a6fa close up view of photon Steps: 64, Sampler: k_euler_a, CFG scale: 30, Seed: 3223332440, Size: 512x512, Model hash: 7460a6fa
|
# ? Dec 31, 2022 01:56 |
|
Swedish Thaumocracy posted:My internet connection died so I asked my locally run copy of stable diffusion to paint me a new one. one of the earliest thing I got I still enjoy looking at AARD VARKMAN posted:"a place that should not exist, colorful photograph by Michal Karcz and Gregory Crewdson" and these AARD VARKMAN posted:"John Goodman swimming in the ocean depths, cosmic horror painting, elegant intricate digital painting artstation concept art by Mark Brooks and Brad Kunkle detailed" -H 704 -C 10.0 -n 9 -s 150
|
# ? Dec 31, 2022 02:00 |
|
Swedish Thaumocracy posted:What are some of the most striking images you lot have made since this whole thing started? i pulled this out of midjourney way back in august: it now hangs on my wall as a glass print
|
# ? Dec 31, 2022 02:27 |
|
Megazver posted:If you make it, I'll come. I have made the thread AI Tools For Traditional Games https://forums.somethingawful.com/showthread.php?threadid=4020822
|
# ? Dec 31, 2022 02:58 |
|
noice It's 4AM here and I'm going to bed, but I'll start posting there tomorrow.
|
# ? Dec 31, 2022 03:00 |
|
https://twitter.com/TheOnion/status/1608896981006487552 img2img, "cyberpunk cyborg criminal with an eyepatch" with a few different models and a couple minor prompt varations
|
# ? Dec 31, 2022 03:23 |
|
Striking? Riddler hideout in Gotham City Batman Picasso's Blue Period Hieronymous Bosch's Times Square And what shoulder and what art Could twist the sinews of thy heart And when thy heart began to beat What dread hand and what dread feet
|
# ? Dec 31, 2022 04:35 |
|
Match these hot new series with their Appropriate Channels ABC Family Discovery Disney Hallmark HBO Syfy PBS 1 2 3 4 5 6 7 Free space: Cartoon network
|
# ? Dec 31, 2022 05:15 |
|
|
# ? Dec 31, 2022 05:21 |
|
Humbug Scoolbus posted:Striking? Thats the stuff.
|
# ? Dec 31, 2022 06:26 |
|
A wizard casting "Prismatic Spray"
|
# ? Dec 31, 2022 07:07 |
|
Megazver posted:How would you prompt this angle to an AI?
|
# ? Dec 31, 2022 08:38 |
|
Don't know if this is the place to ask this but does anyone know if there are image to text AI generators anywhere?
|
# ? Dec 31, 2022 11:47 |
|
You mean something like the CLIP interrogator or something that would try to tell the story behind the picture?
|
# ? Dec 31, 2022 12:12 |
|
Assuming you mean something that will describe a picture for you, the Automatic1111 webui has one built in (CLIP interrogator as stated above), it's geared towards producing prompts that will produce a similar pic, which means it's not great. edit: You can try it out here as well. Swedish Thaumocracy posted:What are some of the most striking images you lot have made since this whole thing started? For all the cool elaborate artsy poo poo I've done, the most striking ones are super simple. Prompt: "Rainbow Lion, Street Art" (imperfectly edited to give it a plain background), I don't have the settings but iirc I used Stable Diffusion 1.5 and the euler scheduler. and this one I printed out on stickers and put it in a few places. It cheers me up every time I see it. Prompt: "The World's Happiest Little Rainbow Dog", probably Waifu Diffusion 1.3 or the NAI leak. The Sausages fucked around with this message at 12:44 on Dec 31, 2022 |
# ? Dec 31, 2022 12:20 |
|
Thanks peeps! Yeah CLIP Interrogator was exactly what I was thinking of.
|
# ? Dec 31, 2022 12:27 |
|
I needed some cozy images to cheer me up KillHour fucked around with this message at 12:40 on Dec 31, 2022 |
# ? Dec 31, 2022 12:33 |
|
It being geared towards producing prompts makes it godawful in a lot of ways, but it's useful for using img2img. because according to CLIP interrogator v2 those last two were "a colorful lion head on a white background, by Lisa Frank, full of colour 8-w 1024, avatar image, light show, platon" and "a dog sitting in the grass with a rainbow in the background, a raytraced image, deviantart contest winner, furry art, cute corgi, 3d model of a japanese mascot, happy sunny day, realistic cartoon" Putting that into SD 2.1 goes about as well as expected: You're overthinking things, CLIP interrogator. edit: goddamn those are cute cozy owls.
|
# ? Dec 31, 2022 12:34 |
|
Actually I'm starting to see all the problems with the cute cozy owls. As well as the concept in the first place. Owls at rest. It's the rest of the loving owl, isn't it? I tried and failed to fix the talons so just got rid of them, and replaced that weird tucked in bit with more feathers. I wonder what else could be improved: I hope that AI artists start to do some editing or reprocessing to fix the details. Sure hands or claws or whatever sucks when AI does it. But it's not that hard to either edit out extra fingers/claws and smooth things out with img2img, or use inpainting to work on problem features until they aren't a problem any more. Maybe 95% of AI art is unusable because of the flaws in the details, and maybe 90% of that is fixable. Apart from AI pics I've only hosed around with lovely memes in GIMP so it's not like it's hard or requires skill to tidy things up a little. My biggest fear about AI art is that the standards are lowered and this sort of poo poo becomes the new normal for the corporate & promo "art" that's shoved in our faces 24/7. AI art is uncanny enough that if it starts appearing everywhere it's gonna cause problems for people with an already loosened grasp on reality. The Sausages fucked around with this message at 13:32 on Dec 31, 2022 |
# ? Dec 31, 2022 13:27 |
|
The Sausages posted:It being geared towards producing prompts makes it godawful in a lot of ways, but it's useful for using img2img.
|
# ? Dec 31, 2022 13:48 |
|
the CLIP stuff is actually how they trained stable diffusion. They first made a model to describe images, then grabbed every image off the internet they could get their hands on and used the AI generated description* to train Stable Diffusion. The anime models are so powerful because Danbooru is just a huge source of correctly tagged images. *Combined with some website-specific information, when available. That's where the whole "trending on Artstation" stuff comes from. Tamba fucked around with this message at 14:10 on Dec 31, 2022 |
# ? Dec 31, 2022 14:02 |
|
baronmasochist posted:Have some knock-off muppets! These were stolen with Cinematic Diffusion and Analog Diffusion. Adding bisexual lighting to prompts with these models really brings out the colors. I haven't used these checkpoints. Did you use them on their own or create merged checkpoints w/SD?
|
# ? Dec 31, 2022 14:49 |
|
Rutibex posted:A wizard casting "Prismatic Spray" I can see a bunch of these on t-shirts.
|
# ? Dec 31, 2022 17:38 |
|
Trent really went out of his way to help us in the future wars, but it's taken a toll on him "Hypnotic sound of silence, echoing through the trees The cracking of the rifles, the marching of the feet I see your world on fire..." quote:syberart trent reznor in SWAT gear and carrying an assault rifle, leaping through a glass window into a burning building, cinematic, highly detailed, epic lighting, realistic war, 4k octane render quote:syberart trent reznor in SWAT gear and carrying an assault rifle, leaping through a glass window into a burning building, cinematic, highly detailed, bisexual lighting, realistic war, 4k octane render
|
# ? Dec 31, 2022 18:15 |
|
I wasn't paying attention for a while, I see there's now checkpoint 1.5 and a whole new release 2.1 Is this the one to use nowadays? https://huggingface.co/stabilityai/stable-diffusion-2-1/blob/main/v2-1_768-ema-pruned.ckpt
|
# ? Dec 31, 2022 18:25 |
|
mobby_6kl posted:I wasn't paying attention for a while, I see there's now checkpoint 1.5 and a whole new release 2.1 Is this the one to use nowadays? https://huggingface.co/stabilityai/stable-diffusion-2-1/blob/main/v2-1_768-ema-pruned.ckpt depends what you want to do. If you're still wanting things in `artgerm` style or wanting super realistic Celebrity Person, or NSFW anything, you want 1.5 still, probably, or some of the other checkpoints based on it. https://www.assemblyai.com/blog/stable-diffusion-1-vs-2-what-you-need-to-know/ 2.x uses a different prompt model where negative prompts are significantly important but also there's a prompt converter on huggingface that will give you a starting point 2.1 positive prompt, based on any 1.5 prompt (so, anything from lexica.art).
|
# ? Dec 31, 2022 18:28 |
A series about diving in the triangle. https://www.youtube.com/watch?v=NXRpuq5LD6A
|
|
# ? Dec 31, 2022 18:41 |
|
accidentally generated the world's most PTSD baby
|
# ? Dec 31, 2022 19:19 |
|
The Sausages posted:Actually I'm starting to see all the problems with the cute cozy owls. As well as the concept in the first place. Owls at rest. It's the rest of the loving owl, isn't it? I'm sorry I didn't open Photoshop and clean up the images I generated while lying in bed unable to sleep to make myself feel better after a really hard day and I just wanted to share them because they made me smile. Dick.
|
# ? Dec 31, 2022 20:04 |
|
I appreciated them, thank you
|
# ? Dec 31, 2022 20:06 |
|
Cabbages and Kings posted:I haven't used these checkpoints. Did you use them on their own or create merged checkpoints w/SD? I don't think I've created a merged checkpoint w/SD yet. I just added the Analog/Cinematic checkpoints to A1111's models directory and switched the model in the web ui.
|
# ? Dec 31, 2022 20:44 |
|
A few random things I've done this month. Prompts are mostly lost to the ages due to inpainting. Blue wizard needs food badly! Trapped in the looking glass Release the robo-geisha
|
# ? Dec 31, 2022 20:57 |
|
KillHour posted:I'm sorry I didn't open Photoshop and clean up the images I generated while lying in bed unable to sleep to make myself feel better after a really hard day and I just wanted to share them because they made me smile. That's fair, I'm genuinely sorry. I'm just venting itt rather than on the platforms where it's an actual problem bc I don't want to touch the poop.
|
# ? Dec 31, 2022 23:33 |
I actually kind of like the images with the errors. Part of what I find interesting about generating A.I art. is seeing where it gets it right and where it gets it wrong. Pieces where people have gone in and corrected mistakes are often "better" but I find the mistakes it makes integral to what makes it interesting to me. You can correct an A.I. image and you have a better image and that's cool and I like it. Or you can not correct an A.I. image and help map the A.I.'s mental landscape and that's cool and I like it.
|
|
# ? Dec 31, 2022 23:43 |
|
The Sausages posted:That's fair, I'm genuinely sorry. I'm just venting itt rather than on the platforms where it's an actual problem bc I don't want to touch the poop. I am not specifically singling you out, but that does bring up a point... It's probably best not to offer critique unless it's explicitly asked for. I believe that was a cardinal rule in the Creative Convention art sharing threads back in the day. No idea if it still is, but it's a good philosophy to follow. Remember that a lot of people are still starting out with this stuff, and some are sharing things they're proud of regardless of any "flaws" you might see. And sometimes those flaws were left in by choice. I occasionally prefer certain imperfections in my images if they enhance the vision I was going for. Pointing out problems discourages others from posting their art here, and I'd rather suffer the occasional seven-fingered hand than have people be afraid to post - and as a result of sharing potentially be inspired to practice and improve.
|
# ? Jan 1, 2023 00:22 |
|
The Sausages posted:But it's not that hard to either edit out extra fingers/claws and smooth things out with img2img, or use inpainting to work on problem features until they aren't a problem any more. you're right and I do this to the 5% of images I care enough about to polish but at coming off high and mighty about people posting poo poo ITT that made them happy that they didn't want to extensively post process. ffs it's also not that hard to think for 10 seconds before posting? 10...9.... ah gently caress it
|
# ? Jan 1, 2023 00:23 |
|
guys did you hear that people are posting unpolished images? And I don't just mean here, I mean in serious Art parts of the internet! edit: quote:syberart artist making unfinished painting on computer, bedroom background, extremely detailed, artgerm 4k hdr Cabbages and VHS fucked around with this message at 00:44 on Jan 1, 2023 |
# ? Jan 1, 2023 00:24 |
|
Alright they apologized and I was pretty grumpy when I replied too so we can probably drop the bag of feathers Something Awful forums member being punished, historical, reproduction KillHour fucked around with this message at 00:56 on Jan 1, 2023 |
# ? Jan 1, 2023 00:50 |
|
|
# ? May 29, 2024 23:09 |
|
here's a couple more cute irradiated mutant babies
|
# ? Jan 1, 2023 00:55 |