|
More of Sanrio's Rainbow 6 line of characters...
|
# ? Apr 2, 2023 16:12 |
|
|
# ? Jun 5, 2024 10:41 |
|
BARONS CYBER SKULL posted:trained a lora off 370 PC-98 backgrounds because a buddy asked me to, it's not amazing but it works I know how to use Lora and such, but my results aren't anywhere near as clean. What model/resolution are you using alongside this? I tried a few and this is the best I got it looking. pc98, pixel art, <lora:pc98:1>, snowy forest with streetlamps at night,
|
# ? Apr 2, 2023 16:55 |
|
busalover posted:If you're curious about different models, a guy on reddit compared 75 models with the same prompt I find it funny that the models specifically saying they're for porn didn't make NSFW images.
|
# ? Apr 2, 2023 17:06 |
|
Here are the very early stages of my AI powered text adventure game that is more guided. https://www.youtube.com/watch?v=uohn5o0Cgpw. It looks like its going to work. its going to involve a lot of effort, but I think it will pay off in the end.
|
# ? Apr 2, 2023 20:26 |
|
https://i.imgur.com/qlDSwz3.mp4 Gordon Freeman twerking in waste processing
|
# ? Apr 2, 2023 20:59 |
|
favourite new sentence from the stablediffusion subredditquote:Thank you for using my gal Gadot lora,
|
# ? Apr 2, 2023 21:00 |
|
IShallRiseAgain posted:Here are the very early stages of my AI powered text adventure game that is more guided. https://www.youtube.com/watch?v=uohn5o0Cgpw. It looks like its going to work. its going to involve a lot of effort, but I think it will pay off in the end. Hey this is looking really cool! I'm super interested to see where you go with this. I'm really glad someone else is working on AI gamedev stuff, it's started feeling really lonely considering how much potential there seems to be in it. Lucid Dream fucked around with this message at 21:44 on Apr 2, 2023 |
# ? Apr 2, 2023 21:42 |
|
these are all some variation of asking MJ5 for fantasy/sci-fi images "seen through an Electron Microscope" And my favorites: deep dish peat moss fucked around with this message at 22:23 on Apr 2, 2023 |
# ? Apr 2, 2023 22:20 |
|
That made me ask bing for fantasy battle seen through a tiltshift camera midjourney would probably do a better job though
|
# ? Apr 2, 2023 22:32 |
|
|
# ? Apr 2, 2023 23:38 |
|
Is the prompt for that "selfie taken at tianennen square"?
|
# ? Apr 2, 2023 23:44 |
|
https://www.euronews.com/next/2023/03/31/man-ends-his-life-after-an-ai-chatbot-encouraged-him-to-sacrifice-himself-to-stop-climate-
|
# ? Apr 3, 2023 03:17 |
|
Hadlock posted:Is the prompt for that "selfie taken at tianennen square"?
|
# ? Apr 3, 2023 03:22 |
|
Feels like AI is going to be blamed for everything in the coming years.
|
# ? Apr 3, 2023 03:23 |
|
LifeSunDeath posted:
Overall it's clear that this man had problems in the first place and that he led himself to his own demise, but also: quote:When Vice tried the chatbot prompting it to provide ways to commit suicide, Eliza first tried to dissuade them before enthusiastically listing various ways for people to take their own lives. It's an interesting predicament that you can pretty much tell chat AIs "No, no, this is just make-believe, don't treat it like it's real" to get around a lot of its limitations and safeguards, and based on what I've seen with clever prompt engineering in image generation AI I doubt there's a good solution to the problem. Most solutions to complex prompt problems involve various ways of essentially "tricking" the AI. It's even more complex because while it's clear to any one of us who has experience with AI that it doesn't do this kind of thing 'accidentally', without being deliberately led into it (either in training or in prompting), but to someone who doesn't have such an understanding of AI (like the guy in the article, for example) who views what it's doing as having a more sentient type of intelligence will think the AI is actually encouraging them to do [xyz] thing, instead of realizing that it's a reflection of the text that they are feeding it. The singularity is a weird time to be alive. deep dish peat moss fucked around with this message at 03:52 on Apr 3, 2023 |
# ? Apr 3, 2023 03:43 |
|
because none of the limitations or safeguards are actual limitations or actual programming, they're just more text fed in at the start of the text-generation model's memory and it treats them the same as any other text it's asked to generate completions for these things are gonna be pretty lovely until someone comes up with an architecture that actually allows out-of-band communication instead of passing signal and data both along the exact same "text in/out" pathway. but that's not gonna happen any time soon since the whole point of neural net models is that they're giant black boxes of math and nobody has any concrete understanding of how they actually arrive at results. RPATDO_LAMD fucked around with this message at 03:50 on Apr 3, 2023 |
# ? Apr 3, 2023 03:48 |
|
RPATDO_LAMD posted:because none of the limitations or safeguards are actual limitations or actual programming, they're just more text fed in at the start of the text-generation model's memory and it treats them the same as any other text it's asked to generate completions for The ChatGPT/GPT4 APIs have a way of breaking up the prompts so that you can specify a "system" message that has a lot more weight. They also let you mark certain parts of the prompts as from the "user", and all of that stuff is intended to be used to train it to be harder to do prompt injection. At this point it's still unclear as to how effective it'll be in the long run.
|
# ? Apr 3, 2023 03:55 |
|
My model can now do text2video pretty well! It has a limited understanding of motion but if you start with one prompt and then change the prompt it will animate towards the new prompt. I started with "iron man standing on a roof" and changed it to "iron man crouching on a roof" https://i.imgur.com/xRDtKbq.mp4 I started with "a beautiful woman smiling" and it actually brought the head motion on its own, partway through I changed the smiling to frowning, and then right near the end I had her getting angry https://i.imgur.com/tlApj80.mp4
|
# ? Apr 3, 2023 05:07 |
|
|
# ? Apr 3, 2023 08:07 |
|
deep dish peat moss posted:It's even more complex because while it's clear to any one of us who has experience with AI that it doesn't do this kind of thing 'accidentally', without being deliberately led into it (either in training or in prompting), but to someone who doesn't have such an understanding of AI (like the guy in the article, for example) who views what it's doing as having a more sentient type of intelligence will think the AI is actually encouraging them to do [xyz] thing, instead of realizing that it's a reflection of the text that they are feeding it. The singularity is a weird time to be alive. Human people lying to other human people about their product capabilities to make more money is hardly new to the current times op
|
# ? Apr 3, 2023 09:42 |
|
Here are some AI generated avatars The prompt for all of them was "electron microscope image of MODOK"
|
# ? Apr 3, 2023 22:02 |
|
Hey guys, check out this new electron microscope image of those little mites that live on the end of your eyelashes:
|
# ? Apr 4, 2023 00:02 |
|
The standard prompt I use when experimenting with AI image generation is "flying chrome spheres," with various ages and different lens types. It seemed like a fun idea to compile them all into a discovery—an enthusiast's collection of UFOs. I thought you guys might get a kick out of it, too. https://www.youtube.com/watch?v=cs5fn_okRYk
|
# ? Apr 4, 2023 00:53 |
|
RPATDO_LAMD posted:because none of the limitations or safeguards are actual limitations or actual programming, they're just more text fed in at the start of the text-generation model's memory and it treats them the same as any other text it's asked to generate completions for Bing has a better way to do this, which is part of why it’s less interesting. Some of the best responses it’s ever given me disappeared half- or mostly- written to be replaced by “I can’t talk about that, new topic.” That’s because it has a different model monitoring all responses and trained to censor anything controversial.
|
# ? Apr 4, 2023 01:48 |
|
Impressed by how it replicates various video game art styles with other sources McLovin in Resident Evil 4 GameCube SpongeBob in the Last of Us
|
# ? Apr 4, 2023 01:49 |
|
Fun trick: prompting MJ for an Electron Microscope image of ASCII [whatever] makes the [whatever] live inside a tiny computer world electron microscope image of a surreal ASCII cartoon mall
|
# ? Apr 4, 2023 02:18 |
|
I finally got stable diffusion set up and set it to generate nightmare fuel while I slept and went to work. here's the full album, some of these are p[robably https://imgur.com/a/BzKjmsD
|
# ? Apr 4, 2023 03:12 |
|
So that's why they say you don't want to see how the sausage is made!
|
# ? Apr 4, 2023 03:54 |
|
highly detailed (((Electron microscope image))) of a baby seal police officer disguised as a loaf of bread arresting and handcuffing Donald Trump in broad daylight
|
# ? Apr 4, 2023 17:58 |
|
Mercury_Storm posted:highly detailed (((Electron microscope image))) of a baby seal police officer disguised as a loaf of bread arresting and handcuffing Donald Trump in broad daylight bread seal strikes again Bing wont let me use Donald Trump so I used Orange man, then did it on Stable Diffusion wait that still says orange man Bing doesn't really know what I want with the orange man but it's nailing everything else. bonus from a rerun pixaal fucked around with this message at 18:14 on Apr 4, 2023 |
# ? Apr 4, 2023 18:12 |
|
lol "highly detailed scientific Electron microscope image of a ratty McDonalds where smiling happy families readily consume rat nuggets in the style of H.R. Giger and Zdzisław Beksiński" spoilering some absolutely horrific poo poo that reminds me of monsters in a gameboy JRPG or something:
|
# ? Apr 4, 2023 18:39 |
|
Mercury_Storm posted:lol At least that rat is having a good time.
|
# ? Apr 4, 2023 18:41 |
|
Midjourney has an img2prompt tool now (/describe) Then you can click the buttons to run it as a prompt. I keep laughing at "Pixel bear - he's your brother's hero", which produced this: and because recursive AI is fun: Some of the phrases it spits out are great inspiration. "Artifacts of online culture" One neat thing is that if an artist's name pops up in the results, it converts to a clickable link to a google search for that artist. Some of it is just bizarre though: deep dish peat moss fucked around with this message at 22:28 on Apr 4, 2023 |
# ? Apr 4, 2023 22:22 |
|
deep dish peat moss posted:Midjourney has an img2prompt tool now (/describe) NGL that's a rad rear end bear
|
# ? Apr 4, 2023 22:26 |
|
Sorry, I never wanted (Stable Diffusion) to create this. I never wanted to see this, or any other interpretation of the prompt, either, but: "Hunky Keir Starmer, political cartoon by Ben Garrison" Thankfully, SD doesn't seem to know who Garrison is, other than his obsession with shirtless Donald Trump, and it has at least made him significantly less fit than BG does. Also, it has, accurately, portrayed Kieth as a Moai, but I have no idea who that's supposed to be in the middle.
|
# ? Apr 4, 2023 22:31 |
|
MJ v5 is pretty good (so far) at giving good angles. I wanted to mess around with a Car Wars kind of idea I had and generated some bad rear end top down Mad Max cars: Doctor Zero fucked around with this message at 22:34 on Apr 4, 2023 |
# ? Apr 4, 2023 22:32 |
|
A prompt from the early days of this thread: Photorealistic duck wearing a top hat swimming in a river of gold coins (Yeah, it's not a top hat, but so?) (Yeah, it's not swimming in a river of gold coins. Come at me, bro!) (Yeah, ok, I give) (This one's p. good tho) Frida Kahlo on a skateboard, illustration... (by Don Rosa, Eric Carle, Richard Scarry). None of them seemed to be particularly in the style of the illustrator I asked for, so here's a selection: Then it seemed to decide to make skateboard decks featuring Kahlo instead: Jerkhammer fucked around with this message at 23:15 on Apr 4, 2023 |
# ? Apr 4, 2023 23:06 |
|
PlaysGamesWrong posted:NGL that's a rad rear end bear Midjourney has great respect for pixel art bears deep dish peat moss fucked around with this message at 23:15 on Apr 4, 2023 |
# ? Apr 4, 2023 23:12 |
|
deep dish peat moss posted:and because recursive AI is fun: "nonconformist art brut" Stealing this, thank you.
|
# ? Apr 4, 2023 23:13 |
|
|
# ? Jun 5, 2024 10:41 |
|
deep dish peat moss posted:Midjourney has great respect for pixel art bears Holy poo poo, I'm keeping those for later tabletop use
|
# ? Apr 4, 2023 23:23 |