|
I can't sleep so please enjoy these urinal birthday cakes.
|
# ? Aug 15, 2022 10:07 |
|
|
# ? May 28, 2024 02:01 |
|
Various things/people as a creepy monster in your closest courtesy of stable diffusion A hamburger A pizza A grey alien Shaggy
|
# ? Aug 15, 2022 22:13 |
|
https://twitter.com/weirddalle/status/1559085066151952385
|
# ? Aug 16, 2022 04:16 |
|
I've gotten Dall-E 2 access. Photograph of master chef making a urinal cake Photograph of Steamed Hams A still from the pokemon anime where an ancient evil lovecraftian entity is released Still of Neil Breen in Dune (1984) A taxidermied pikachu Tommy Wiseau in the videogame Silent Hill 4 The Room Photograph of Cryptid spotted in McDonalds (1996) A still of Danny Devito in Bloodborne (2015) A still from the film Bloodborne (2015) Still of columbo in Bloodborne (2015) Photograph of eldritch horror working at Wal-Mart A still of Elmo in A Clockwork Orange (1971) Still of Guybrush Threepwood in Pirates of the Caribbean (2003) A still of Kermit the Frog in Vampire Hunter D Bloodlust (2000) Photograph of a Vampire at a Disco club during the 1970s Still of Shrek in Avatar (2009)
|
# ? Aug 16, 2022 05:18 |
|
IShallRiseAgain posted:Still of Guybrush Threepwood in Pirates of the Caribbean (2003) I can't tell whether I love or hate it
|
# ? Aug 16, 2022 05:21 |
|
Ok I think I'm starting to agree with the cranks on this https://mobile.twitter.com/adamszym/status/1557497131044331526
|
# ? Aug 16, 2022 15:41 |
|
Agreeing with joe football, Stable Diffusion is shockingly good at celebrity recreations. I finally started playing with it earlier this week. I don't know why, but I got addicted to the prompt "<x> working the front counter of mcdonald's ready to take your order, eighties polaroid, flash photography, photorealistic, kodachrome" replacing <x> with different celebrities. Anya Taylor-Joy Alex Trebek Dolly Parton DJ Khaled Selena Gomez Jason Voorhees Scarlett Johansson Hugh Jackman as Wolverine Albert Einstein Nayeon The Rock with an afro and a tight black turtleneck and fannypack Solid Snake (David Hayter) Doja Cat Kermit the Frog I also agree with the assessment that it's great at remixing items, concepts, and people it knows in its database, but it has a huge amount of difficulty creating something it doesn't seem to know about. The above celebrities worked, but I've had trouble with things like Optimus Prime, The Terminator, and the Gorn captain from Star Trek. Elmo turned out to be a guy in a red rooster costume. I put a member of a k-pop group in the mix above. Strangely enough, the database absolutely knows about her, but for some reason biffs it on any other member of the same group. Trying any other member's name results in a weird amalgamation of members of that group, as if it only had group photos to base its likeness of the character on.
|
# ? Aug 17, 2022 06:24 |
|
Kennel posted:Did you mean to post that to Waffleimages thread or something? Belated but yeah sorry, had both threads open at the time
|
# ? Aug 17, 2022 06:29 |
|
|
# ? Aug 18, 2022 03:42 |
|
jason with his own mask on his shirt
|
# ? Aug 18, 2022 03:57 |
|
quote:A cow...? In Africa...?
|
# ? Aug 19, 2022 00:00 |
|
|
# ? Aug 19, 2022 00:36 |
|
|
# ? Aug 19, 2022 17:59 |
|
https://twitter.com/ai_curio_bot/status/1560600892579098627
|
# ? Aug 19, 2022 18:17 |
|
https://twitter.com/kennykeil/status/1559935527897288704
|
# ? Aug 19, 2022 23:40 |
|
https://twitter.com/GabriellaG439/status/1561207676696027140 https://twitter.com/GabriellaG439/status/1561420578178809856
|
# ? Aug 22, 2022 00:44 |
|
More: https://www.reddit.com/r/SubSimulatorGPT2/comments/wu1oue/kermit_the_frog_vs_miley_cyrus/
|
# ? Aug 22, 2022 08:47 |
|
multiversus gettin real weird
|
# ? Aug 22, 2022 09:07 |
|
https://twitter.com/EMostaque/status/1561777122082824192 SD open source released!
|
# ? Aug 22, 2022 19:12 |
|
Tunicate posted:SD open source released! And a web-based generator is already available again: https://huggingface.co/spaces/stabilityai/stable-diffusion Not as fast at churning out images as the discord bot, but it's still good, and the price is absolutely right. "a corgi wearing an astronaut suit, dragoncon, 35mm, f/2.8, flash fired, full length"
|
# ? Aug 23, 2022 03:19 |
|
https://twitter.com/kushaanshah/status/1561771338510921728
|
# ? Aug 23, 2022 04:26 |
|
|
# ? Aug 23, 2022 17:34 |
|
Looks like all the off-brand movie poster artists are gonna be out of a job
|
# ? Aug 23, 2022 17:47 |
|
This is what happened when I tried to get Stable Diffusion to show me a presidential debate between Pikachu and Chewbacca
|
# ? Aug 23, 2022 18:18 |
|
Doggles posted:And a web-based generator is already available again: I'm at 1900 seconds and still going on my first prompt lol
|
# ? Aug 23, 2022 18:20 |
|
AARD VARKMAN posted:I'm at 1900 seconds and still going on my first prompt lol You might want to try again. I've only done a few so far, but it seems to vary a lot. I haven't seen it go much over 500 though.
|
# ? Aug 23, 2022 18:24 |
|
END CHEMTRAILS NOW posted:Out of how many? It should tell you the total time and how much progress you've made, like "Queue: 11/16 | 159.4/362.6" worked the second time, I think the first time it got hung up on processing in a loop or something - I got to the end of the queue and then it just kept going forever
|
# ? Aug 23, 2022 18:59 |
|
Doggles posted:And a web-based generator is already available again:
|
# ? Aug 24, 2022 01:25 |
|
Interestingly, you don't actually need a prompt to have Stable Diffusion generate content. edited to add: I'd heard a casting rumor and wanted to visualize it, but this kind of seems like it decided to make Esposito both a professor and Malcolm X. (further edit, my girlfriend has just informed me that Giancarlo Esposito wears glasses, so maybe not anything to do with Malcolm X) obsolete absolution has a new favorite as of 03:10 on Aug 24, 2022 |
# ? Aug 24, 2022 02:54 |
|
obsolete absolution posted:edited to add: Edit: And now it just loaded. Weird.
|
# ? Aug 24, 2022 03:07 |
|
"Bernie Sanders shirtless with a rocket launcher"
|
# ? Aug 24, 2022 04:28 |
|
Ok who gave AIs access to the topless Bernie training data
|
# ? Aug 24, 2022 04:38 |
|
obsolete absolution posted:Interestingly, you don't actually need a prompt to have Stable Diffusion generate content. I think you do if the guidance_scale parameter is 1? Not a hundred percent on that.
|
# ? Aug 24, 2022 05:38 |
|
|
# ? Aug 24, 2022 06:14 |
|
I think they've seen a few too many bombed out humvees, like GIP.
|
# ? Aug 24, 2022 06:26 |
|
Doggles posted:And a web-based generator is already available again: Some 1:1 dall-e prompt comparisons: (SD probably needs different formatting for good results I guess?) Exterior othographic isometric scene of a Capsule Corp tower in a lush flower field :: Style: Akira Toriyama, kurzegasgt, Studio Ghibli, Digital Art dall-e: stable diffusion: Exterior. View from penthouse balcony. Hyperactive megacity laced with neon, advertisements, street market, pedestrians :: Style: Shadowrun, Cyberpunk, Scifi, Futuristic, Digital Art dall-e stable diffusion: stable diffusion without the style tags ("Exterior view from penthouse balcony. Hyperactive megacity laced with neon, advertisements, street market, pedestrians") Without the style tags it's just sort of a slightly improved version of old 2019-2021 era AI Art Gen. I didn't try e.g. LatentVisions or any of that older stuff with style tags and things like that though. (not wasting a dall-e credit on this b/c it would look bad) Digitalillustration by Alex Ross and HR Giger of a gigantic eldritch being kneeling to inspect a suburban house at night. An eerie eldritch glow in the eerie fog. Toonart. dall-e stable diffusion my takeaway at first glance: SD is better at fine details and thus it's better at understanding the shapes and lines between things and has more clearly defined shapes and objects. Stylistically they are entirely different beasts, and content-wise it seems like SD tries to give you literally what you asked for whereas dall-e interprets it artistically (dall-e essentially starts as a 64x64 pixel grid and each pixel of that grid ends up representing a 16x16 chunk of the final image so it's not great about fine detail at a scale smaller than 16x16, vs SD which it seems works on smaller scales) SD feels like a linear evolution of 2019-2021-era ai art gen like LatentVisions and all that old Deep Dream-based stuff where it's doing the same thing but in higher definition, vs Dall-E which much more elegantly processes language and combines concepts together. SD is the boring but detail-oriented technical artist and dall-e is the imaginative concept artist. Really jazzed about the way SD interpreted a combination of Alex Ross and Giger though. If it understood the contents better that'd be a rad illustration style re: SD's lack of creativity - I ran the same prompt again and one of the generations was almost identical to one of the generations from the first run: Which reinforces my "SD = 'boring' but technical, visual-detail-oriented and reliable, dall-e = wild and imaginative but not so great at delivering fine detail." assessment. Both are cool tools, Dall-E has way more potential than any of the others I've seen though. I blew through over 2000 credits figuring things out with it and it's truly remarkable when it comes to developing distinct non-derivative visual styles that can be repeated across multiple prompts. Also if anyone is curious here's a huge dump of dall-e generations, not even close to everything I generated but: Sci-fi cityscapes: https://www.behance.net/gallery/151053673/AI-Generated-Sci-Fi-Cityscapes-Ben-X-DALL-E Eldritch lovecraftian nightmare world: https://www.behance.net/gallery/151054507/AI-Generated-Artwork-Eldritch-Lovecraftian-Nightmares Character and monster portraits/designs: https://www.behance.net/gallery/151055431/AI-Generated-Art-CharacterMonster-Concepts Isometric scenes: https://www.behance.net/gallery/151055767/AI-Generated-Art-Isometric-Scenes Random planetscapes: https://www.behance.net/gallery/151056063/AI-Generated-Art-Planetscapes Psychedelic/Cosmic Pop Art: https://www.behance.net/gallery/151056495/AI-Generated-Art-CosmicPsychedelic-Pop-Art Various landscapes: https://www.behance.net/gallery/151057111/AI-generated-Art-various-landscapes deep dish peat moss has a new favorite as of 07:31 on Aug 24, 2022 |
# ? Aug 24, 2022 06:34 |
|
What's your vote?
|
# ? Aug 24, 2022 18:20 |
|
doomguy ahegao
|
# ? Aug 24, 2022 18:25 |
|
Indiana Jones surrounded by woodland creatures like a Disney princess. And a portrait of my current TTRPG character. If I was half-way competent at visual art type stuff I'd be tempted to tweak a few things, but I'm not so I'm just going to use it as is because I think it came out pretty good.
|
# ? Aug 24, 2022 18:53 |
|
|
# ? May 28, 2024 02:01 |
|
I love that the AI interpreted the prompt as though “Disney Princess” was an example of a woodland creature. Beautiful example of a computer doing what you tell it, not what you mean.
|
# ? Aug 24, 2022 19:16 |