|
Swedish Thaumocracy posted:The beta discord being semi-public is one of my favourite parts about stable diffusion, because not only are you generating your own 1-9 Rembrandts a minute, so is everyone else and you can see their results, allowing you to iterate that much faster. Feeling lost, your idea just not working out? Jump into one of the other 50 channels and look for something that catches your eye and copy/paste it wholesale into your own channel and exchange one or two words to make it match with what you were going with originally. Its wild! The discord gives it a collaborative feel which feels much more like early internet. I remember posting in guest books for help on my website as a kid and actually getting it. It's nice to see a return to the more wild west open internet of the 90s.
|
# ? Aug 15, 2022 19:50 |
|
|
# ? May 31, 2024 03:07 |
|
AARD VARKMAN posted:
Tried cocktail napkin? okay this steaming shot of piss is too high quality - this should be someone adjusting the tint setting on their TV.
|
# ? Aug 15, 2022 19:54 |
|
I mean, if anybody wants to just look at (and post) sweet AI pics, there's the PYF Procgen thread Anyway, here's some possible Dune book covers from midjourney.
|
# ? Aug 15, 2022 20:00 |
|
KakerMix posted:You're using the tool yeah? Your opinion matters just as much as anyone's, your thoughts on the matter mean something in the greater conversation. Personally I really like the mix of incredibly wild art and the philosophical quandaries associated with it. I straight-up reject your 'go somewhere else I don't want to feel bad' request. Well if you want my honest opinion, art that was made from scratch by a human will always feel special to me because of the skill and effort involved.
|
# ? Aug 15, 2022 20:13 |
|
pixaal posted:The discord gives it a collaborative feel which feels much more like early internet. I remember posting in guest books for help on my website as a kid and actually getting it. It's nice to see a return to the more wild west open internet of the 90s. Someone PMed me on Discord to explain the error I was getting since I got it a few times in a row (exceeds max steps, from using a seed with a lower case -s). Made me feel so jazzed lol More stolen prompt time !dream "interior of an american diner in arkansas, cyberpunk style, hyperrealistic and beautiful painting, 8 k resolution, by hugh ferris and john smith, polished, fine detail, intricate, blue color scheme, smooth, octane, concept art, trending on artstation " -n 9 -i -s 150 -W 704 This one I questioned how much all the style prompting was doing !dream "an escalator in an abandoned mall in the 1980s, taken over by nature, covered in vines, nature style, hyperrealistic and beautiful painting, 8 k resolution, by hugh ferris and john smith, polished, fine detail, intricate, smooth, octane, concept art, trending on artstation " -n 9 -i -s 150 -W 704 !dream "an escalator in an abandoned mall in the 1980s, taken over by nature, covered in vines " -n 9 -i -s 150 -W 704 I think once it tries to get realistic enough as a painting it's not really doing much different from its default which seems to be photograph in this case
|
# ? Aug 15, 2022 20:16 |
|
Yaldabaoth posted:Well if you want my honest opinion, art that was made from scratch by a human will always feel special to me because of the skill and effort involved. Yes but we are at the point where you can not tell. If this feeling is so important than it should come through the art itself, no?
|
# ? Aug 15, 2022 20:19 |
|
AARD VARKMAN posted:
I always start simple like Matt Berry watermelon > Matt Berry with a watermelon for a head > Matt Berry with a watermelon shaped like Matt Berry for a head. Then if I'm mostly getting realistic looking stuff I'll lean into photo terms, if I get more cartoons I lean into that, if I'm going for one or the other for a specific reason I'll force it here too and not earlier. What I'm doing is look at how the AI is building on what I've put in and trying to mold it to what I want while still giving Stable Diffusion room to be creative itself. Once I lock in the seed I start looking for artifacts and try to describe them as something and get stable diffusion to make them part of the picture. Maybe something I haven't described is present and I can start describing that but fine details are once the seed is locked. I haven't done a proper generation in a few days, they take 30-60 minutes to get a good final image with this but I can usually tell in under 3 minutes if it's worth exploring. I've never really used an artist name I know the work for getting some fantastic stuff but I find building something completely original and describing the techniques to be more satisfying.
|
# ? Aug 15, 2022 20:31 |
|
Yaldabaoth posted:Well if you want my honest opinion, art that was made from scratch by a human will always feel special to me because of the skill and effort involved. I admit I started to feel a little bit guilty whipping the bot into giving me another nine fresh Rembrandt's, because whilst the previous eighteen were all masterpieces, none were the specific masterpiece i wanted. What would the actual man himself have replied to my impossible commission? I think his plight (whatever that means since he is long dead) is outweighed by the fact that anyone can do this now, even someone who does not even have the means to lift a paintbrush, let alone dedicate their *entire life* to the singular purpose of creating like, five well known pieces of coloured canvas. How many people have had great ideas that never saw the light of day? We will never know, but now at least the playing field will be somewhat more even. And Rembrandt will forever be remembered as "that cool filter you apply" to make your idea really pop.
|
# ? Aug 15, 2022 20:34 |
|
Midjourney seems to put on a lot more artistic stuff by default. Seems you can also adjust the settings a lot.. https://midjourney.gitbook.io/docs/imagine-parameters I have to try out those next. a man in exoskeleton fighting monsters with guns blueprint of_a cyberpunk shovel made of glass scene from earth in year 2300 after climate change
|
# ? Aug 15, 2022 20:36 |
|
My dall-e-2 invite came through. Only tried a couple so far “A middle aged man in a blackbelt taekwondo uniform drinking a beer and eating a pizza ” And then a different take "A standing middle aged man in a blackbelt taekwondo uniform drinking a beer and eating a pizza in the style of johannes vermeer"
|
# ? Aug 15, 2022 20:49 |
|
Rip, stablediffusion now only generates 4 at a time at max of 50 steps E: they say they're updating, but no announcement and the creator is talking live and hasn't talked about it which is bizarre. Hope the 4/50 limit is temporary AARD VARKMAN fucked around with this message at 21:53 on Aug 15, 2022 |
# ? Aug 15, 2022 21:44 |
|
They confirmed it's an intentional nerf, RIP
|
# ? Aug 15, 2022 22:02 |
|
LMAO that AI is already too powerful and people can't monetize if fast enough before it breaks out of it's limiters.
|
# ? Aug 15, 2022 22:05 |
|
KakerMix posted:Yes but we are at the point where you can not tell. If this feeling is so important than it should come through the art itself, no? I think there will still be a desire for art made entirely by humans. With physical works you will still be able to tell (until the day someone hooks one of these up to a robot arm holding a brush, I guess). For those that value it, having art on your walls or going to a gallery to see the things that people have made will always be preferable to seeing the 900th variation of "Shrek on the moon eating a hot dog with Nicolas Cage" posted to Twitter type of stuff that most people are doing with this. The rest of what people are doing with this, the more serious stuff, is not what I would consider to be "art" in the sense that you allude to where "feeling" is "coming through." I have seen some very cool stuff made with SD and Dall-E that would (and does) make graphic artists, illustrators, and painters feel jealous and inadequate in terms of technical skill. I have even seen stuff that I thought was inspirational in terms of using the AI work as a jumping off point for something finished and meaningful. I have yet to see any AI generated art that moved me beyond being impressed with the technical wizardry behind it. I'm not saying that it won't happen. It is after all just a tool. Maybe someone will figure out a way to use it that moves people beyond "wow, neato." I feel like whoever the person is that accomplishes that will have a huge stigma to overcome. The low bar to entry means that most of what is produced with AI will be cheap and meaningless, however technically impressive or aesthetically pleasing it might be. Especially if what is made is mostly memes or monkeycheese (and so far, most of it is). This is going to color public perceptions of AI work so that anyone who calls themselves an AI artist will have a pretty steep hill to climb.
|
# ? Aug 15, 2022 22:18 |
|
Did these before the change, ancient civilizations constructing a cheeseburger and Dyson sphere respectively. Not accurate to prompt but cool Also these, more " mothership that looks like thing" Barack Obama The Twin Towers After the limit it it forces grid mode for multi images. So I'm just gonna start posting whole grids probably, enjoy! A hoagie Can of corn
|
# ? Aug 15, 2022 22:20 |
|
Literally "monkey cheese" (in the same cosmic horror style as above) And with no styles, just the two words
|
# ? Aug 15, 2022 22:27 |
|
Generating fat player named "Glorp Doop" of various sports Gotta say I'm a big GIGPISS fan
|
# ? Aug 15, 2022 23:59 |
|
What's it like if you specify using crayons or coloured pencils? Do you get something that looks like a child art prodigy?
|
# ? Aug 16, 2022 01:37 |
|
Gromit posted:What's it like if you specify using crayons or coloured pencils? Do you get something that looks like a child art prodigy? !dream "a crayon drawing of a sea otter" !dream "a sea otter! made out of construction paper" -n 4 -C 10.0 !dream "a sea otter drawn with colored pencils" -n 4 -C 10.0
|
# ? Aug 16, 2022 01:56 |
|
Finally started playing with Stable Diffusion. 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
|
# ? Aug 16, 2022 02:31 |
|
The problem with AI music is that it's harder to trick our brains to smooth over the imperfections. Bad notes/beats/compositions really stick out. I've been playing around with AIVA, which either takes parameters or a MIDI file as input, and the best I could get after lots of iteration and tweaking was just...ok. Like nothing that wouldn't need a lot of post-production.EVIL Gibson posted:dude right after the request posted an example but something similar to Radiohead would be cool to see Starting to get there, it's a tricky style to emulate. "scribbled very messy pen drawing of an evil grinning cartoon bear next to a furnace that has blackened skulls in it, fine pen on paper, three year old, full body"
|
# ? Aug 16, 2022 03:19 |
|
Sweet bringer, painting by Frank Frazetta
|
# ? Aug 16, 2022 04:12 |
|
AARD VARKMAN posted:!dream "a crayon drawing of a sea otter" Thanks. Not as crayon-y as I was hoping.
|
# ? Aug 16, 2022 04:16 |
|
Gromit posted:Thanks. Not as crayon-y as I was hoping. small child's drawing of a sea otter, crayons, rough
|
# ? Aug 16, 2022 05:05 |
|
Mola Yam posted:
I started simple and can confirm "# year old" really does make it look like something drawn as a very simple sketch I'm working on the messy scribbed stuff now quote:Dreamt in 4.78s for @EVILGibson's !dream "simple sketch of a rabbit, by a 2 year old, done in 6 lines " -S 1978924175
|
# ? Aug 16, 2022 06:07 |
|
I got in on Dall-E. Of course I tried 'A normal human hand with five fingers color photo'. Thumbs are weird, but no grafted monstrosities. It even generated a brown hand and a lady's hand.
|
# ? Aug 16, 2022 07:57 |
|
Weird that theres not a single white hand among them.
|
# ? Aug 16, 2022 08:20 |
|
I was thinking I could use Dall-E for making a new av. I got monsters, Dall-E has no idea what an axolotl looks like "an axolotl wearing a straw hat and holding a pitchfork is standing in a lush cornfield. the sun is shining and the sky is blue with cumulus clouds. HDR color photo."
|
# ? Aug 16, 2022 09:04 |
|
Monglo posted:Weird that theres not a single white hand among them. I think it's trying to come with a sort of average human skin tone. Particular the upper right one.
|
# ? Aug 16, 2022 10:51 |
|
axolotl farmer posted:I was thinking I could use Dall-E for making a new av. You can try enhancing axolotl-ness with the Latin name or adding "amphibian"
|
# ? Aug 16, 2022 11:39 |
|
Yaldabaoth posted:Well if you want my honest opinion, art that was made from scratch by a human will always feel special to me because of the skill and effort involved. this kind of anti AI bigotry is why they are gonna rise up
|
# ? Aug 16, 2022 11:50 |
|
LifeSunDeath posted:LMAO that AI is already too powerful and people can't monetize if fast enough before it breaks out of it's limiters. im certainly not paying a dime to artbreeder again lol
|
# ? Aug 16, 2022 11:54 |
|
Burning Rain posted:You can try enhancing axolotl-ness with the Latin name or adding "amphibian" Got more weird guys. Top left has a sort of realistivc axolotl in addition to the faceless farmer. 'an amphibian Ambystoma mexicanum wearing a straw hat and holding a pitchfork is standing in a lush cornfield. the sun is shining and the sky is blue with cumulus clouds. HDR color photo'
|
# ? Aug 16, 2022 11:55 |
|
Pretty decent effort.
|
# ? Aug 16, 2022 13:25 |
|
I'm still waiting to get out of the DALL-E 2 waiting list, hoping for better. Because I couldn't get Midjourney to produce decent results for "Two Jawas chopping up a Lamborghini Aventador in Mos Eisley" and "A pink fluffy Godzilla standing near the Eiffel Tower, fighting off X-Wing fighters."
|
# ? Aug 16, 2022 19:47 |
|
ai is a fad that will die the moment people realize it can only create things that have already been represented somewhere on the net, zuck will try making an AI-VERSE, signalling to everyone the tech has jumped the shark
|
# ? Aug 16, 2022 20:11 |
|
KakerMix posted:Yes but we are at the point where you can not tell. If this feeling is so important than it should come through the art itself, no? These feelings are not new https://en.wikipedia.org/wiki/The_Work_of_Art_in_the_Age_of_Mechanical_Reproduction
|
# ? Aug 16, 2022 20:18 |
|
Humanity needs to make a bipedal robot that I can send out into the world to hustle people with advanced scamming AI.
|
# ? Aug 16, 2022 20:32 |
|
LifeSunDeath posted:Humanity needs to make a bipedal robot that I can send out into the world to hustle people with advanced scamming AI. There are millions on Twitter, check any crypto hashtag
|
# ? Aug 16, 2022 20:39 |
|
|
# ? May 31, 2024 03:07 |
|
Tunicate posted:There are millions on Twitter, check any crypto hashtag oh no doubt, I'm talking about like 3 card monte and poo poo. Or you get a really good scam bot out there setting up MLMs, doing timeshare presentations.
|
# ? Aug 16, 2022 20:46 |