|
 Wario, portrait by Joel-Peter Witkin
|
#
?
Jun 19, 2022 00:17
|
|
|
|
| # ? May 30, 2024 09:38 |
|
|
|
|
#
?
Jun 19, 2022 02:11
|
|
|
  
|
|
#
?
Jun 19, 2022 03:32
|
|
|
  
|
|
#
?
Jun 19, 2022 03:47
|
|
|
This is stilly my favorite Fieri one: 
|
|
#
?
Jun 19, 2022 04:23
|
|
|
gabe newell eating a hamburger in the style of klimt
|
|
#
?
Jun 19, 2022 05:03
|
|
|
This Is the Zodiac posted:
This is awesome
|
|
#
?
Jun 19, 2022 08:53
|
|
|
Entropic posted:This is stilly my favorite Fieri one: Mad Guy Flavor Road
|
|
#
?
Jun 19, 2022 12:54
|
|
|
This Is the Zodiac posted:
Is that a piranha plant?? Crazy
|
|
#
?
Jun 19, 2022 12:57
|
|
|
 
|
|
#
?
Jun 19, 2022 13:48
|
|
|
strange feelings re Daisy posted:After seeing this I wasted like 20 attempts trying to render a blue-eyed Guy Fieri barbecuing in a Fremen stillsuit. I could get one or two of those things in the image, but never all of them at once. This is my new white whale. Only a very powerful AI can handle this imagery. I just scored an invite to Dall-E 2 and I tried this out with your prompt. Removed the realistic faces one as its against OpenAi's terms to share. Sorry! Specifically:  " blue-eyed Guy Fieri barbecuing in a Fremen stillsuit in the style of anime" OpenAi is avoiding 'real' people as hard as it can which means there is no real 'Guy' here. EDIT ymgve posted:You should note that the OpenAI terms of service forbids sharing pictures with realistic faces, so be careful with where you share those You are right, edited. Thanks for looking out. KakerMix fucked around with this message at 16:09 on Jun 19, 2022 |
|
#
?
Jun 19, 2022 15:36
|
|
|
You should note that the OpenAI terms of service forbids sharing pictures with realistic faces, so be careful with where you share those
|
|
#
?
Jun 19, 2022 15:48
|
|
|
 
|
|
#
?
Jun 19, 2022 15:49
|
|
|
More stuffs that I've gotten out of Disco Diffusion. I like a fair amount of this. I typed in something like "The World is my Playground" by Don Bluth  City being held up by balloons  A haunted house inspired by Mike Mignola  Eyvlind Earle doing the concept art for Fantasy Planet  Encanto being Encanto 
|
|
#
?
Jun 19, 2022 18:02
|
|
|
KakerMix posted:
well, gently caress OpenAI then. I know why they are doing it; to prevent lawsuits against them out the butt but it's sad poo poo to see these clauses.
|
|
#
?
Jun 19, 2022 18:17
|
|
|
 Top center came close but I was hoping for that steady gun Vasquez holds.
|
|
#
?
Jun 19, 2022 18:25
|
|
|
EVIL Gibson posted:well, gently caress OpenAI then. I know why they are doing it; to prevent lawsuits against them out the butt but it's sad poo poo to see these clauses. They are the most popular one for exposure, and currently the most powerful. In a very short amount of time I expect to see a bunch of more powerful and without restrictions AIs being put out that will shame Dall-E. It's too powerful and too 'easy' now that we get how to do it. It doesn't even have to be a profit driven thing either, smart people out there with the resources will make a more powerful Dall-E mini just because they disagree with OpenAI's stance. I hope everyone is really ready for this worming it's way into everything cause it's fun now but growing pains are ahead of us.
|
|
#
?
Jun 19, 2022 18:43
|
|
|
KakerMix posted:They are the most popular one for exposure, and currently the most powerful. In a very short amount of time I expect to see a bunch of more powerful and without restrictions AIs being put out that will shame Dall-E. It's too powerful and too 'easy' now that we get how to do it. It doesn't even have to be a profit driven thing either, smart people out there with the resources will make a more powerful Dall-E mini just because they disagree with OpenAI's stance. like every new media it will explode exponentially once the porno people get ahold of it
|
|
#
?
Jun 19, 2022 18:49
|
|
|
KakerMix posted:They are the most popular one for exposure, and currently the most powerful. In a very short amount of time I expect to see a bunch of more powerful and without restrictions AIs being put out that will shame Dall-E. It's too powerful and too 'easy' now that we get how to do it. It doesn't even have to be a profit driven thing either, smart people out there with the resources will make a more powerful Dall-E mini just because they disagree with OpenAI's stance. Welll... trianing the largest models takes millions of dollars and requires non-consumer-grade hardware, so it's not like Just Some Smart Guy can do it
|
|
#
?
Jun 19, 2022 18:51
|
|
|
Holy lol, I just realized that this may change Art Historian from being a joke degree into one that's extremely sought after since they'll know all the secret codes for art AI inputs
|
|
#
?
Jun 19, 2022 19:18
|
|
|
Wheany posted:Welll... trianing the largest models takes millions of dollars and requires non-consumer-grade hardware, so it's not like Just Some Smart Guy can do it Hence "resources". This isn't gated behind giant companies is what I'm getting at, and this isn't even getting into state-backed initiatives either.
|
|
#
?
Jun 19, 2022 19:20
|
|
|
 Explains why nobody ever comes to visit us, we live in one of the dark patches. We are in the boonies of the network.
|
|
#
?
Jun 19, 2022 19:53
|
|
|
KakerMix posted:
|
|
#
?
Jun 19, 2022 20:02
|
|
|
KakerMix posted:
|
|
#
?
Jun 20, 2022 01:43
|
|
|
I don't want to risk my new toy I'm sure you understand
|
|
#
?
Jun 20, 2022 02:27
|
|
|
https://twitter.com/pattonoswalt/status/1538708722196832256?s=20&t=QMRLA7sKU0oqTmHag24PuQ
|
|
#
?
Jun 20, 2022 03:30
|
|
|
KakerMix posted:Hence "resources". This isn't gated behind giant companies is what I'm getting at, and this isn't even getting into state-backed initiatives either. Pornhub often does gimmicky stuff, and I imagine with their worth ($1.6 billion) they could scrounge up the money to make a pro-porn AI that would probably see insane levels of engagement and ad revenue far beyond the costs to fund it. Even though I am sure they would have to block all porn of *insert name of person or IP*, people being able to create their own fantasy settings and type of sex act and person would still be a huge draw, IMO. Chunk Tingle novels come to life, etc. An Ounce of Gold posted:Encanto being Encanto This looks amazing, yet also looks like the next AAA Breath of the Wild clone.
|
|
#
?
Jun 20, 2022 05:29
|
|
|
  
|
|
#
?
Jun 20, 2022 14:23
|
|
|
Seems fairly accurate  E- 
Snowy fucked around with this message at 14:32 on Jun 20, 2022 |
|
#
?
Jun 20, 2022 14:28
|
|
|
 
|
|
#
?
Jun 20, 2022 17:33
|
|
|
Wrapping up my animation. Going to generate a song to pair with it (which seems like it could take a while to finish), then maybe write up everything I found for anyone else interested in making amazing dreamy art. https://www.youtube.com/watch?v=gmpdOxpKNCE If you're interested in nightmare AI music to accompany your disco creations, this seems to be the right answer: https://robertbrucecarter.com/writing/2020/07/a-simple-openai-jukebox-tutorial-for-non-engineers/ At this point my process is: -Find a cool prompt, set up some z-movement in DD and set the prompt to change slightly every few hundred frames -DD renders frames -Realesrgan to upscale frames -Either back to DD or use another software to interpolate frames (I'm using Premier to double the framerate but half the video speed to interpolate every other frame) This is cool for creating constantly morphing dreamscapes, but really bad at creating a coherent video. There's two other amazing ways to do this I've found: -Similar to what I'm doing in DD, but more complicated prompts. Use a static seed, and use your frame number to progress your prompt (example: A [Frame#] year old woman will make an animation of a woman growing old). You can get really creative with weights and changing up your prompt here, and ultimately you're interpolating concepts in your work. Should be possible to do something like make an animation of a mushroom house with prompts about a mushroom that sprouts then rots, and additional prompts about things like doors, windows and chimneys that increase in weight when the mushroom is 'in its prime'. -Another option is using video input. DD supports an image input it will use as a base image to 'morph', so you can either feed it each frame of a video and give it a prompt, or the latest version supports video input. So people are doing things like feeding it videos of them dancing or claymation with prompts like 'tentacles' to get amazing results. Atoramos fucked around with this message at 19:44 on Jun 20, 2022 |
|
#
?
Jun 20, 2022 18:56
|
|
|
 
|
|
#
?
Jun 20, 2022 19:54
|
|
|
|
|
#
?
Jun 20, 2022 20:01
|
|
|
|
|
#
?
Jun 20, 2022 20:19
|
|
|
Pot pot pie
|
|
#
?
Jun 20, 2022 20:27
|
|
|
I got an invite to Dalle-2 a day after signing up for OpenAI's Playground. I quickly hit the daily limit of 50 prompts just messing around. Renaissance painting of Daft Punk performing a concert   Daft Punk comic cover illustrated by Bill Sienkiewicz    Renaissance painting of Garfield attempting to eat an entire turkey    
|
|
#
?
Jun 20, 2022 21:03
|
|
|
I cannot believe how good the dall-e-2 poo poo is 
|
|
#
?
Jun 20, 2022 21:06
|
|
|
 his eyes 
|
|
#
?
Jun 20, 2022 21:21
|
|
|
|
|
#
?
Jun 20, 2022 21:26
|
|
|
|
| # ? May 30, 2024 09:38 |
|
|
At last I have a place to dump all my stupid dall-e images.                  
|
|
#
?
Jun 20, 2022 21:31
|
|