|
wait, it's okay to do that? i can make a comic with DALLE???
|
#
?
Aug 4, 2022 00:25
|
|
|
|
| # ? May 29, 2024 19:32 |
|
|
Mr Ice Cream Glove posted:What will be interesting to see is if things that are generated by this AI tool get used in media (games,comics,movies,tv,album covers) I'm convinced it's already in media and we just don't know it yet. naem posted:we are seeing nothing less than the automation of intellectual labor. I'd actually be more worried if I were at Disney, they've been so concerned with removing their need for people and it's going to skip right over them and not need them as a corporation. Why, you can just have an AI generate whatever fiction with whatever parameters you want. The very idea of entertainment as a business is at risk here, who needs a conglomerate? If the AI can do everything Disney does, what good is Disney to the end user? fez_machine posted:Abhay Khosla of Abhay Khosla's Bram Stoker's Dracula has been making comics using MidJourney and the results are pretty good. Awesome
|
|
#
?
Aug 4, 2022 00:30
|
|
|
yeah commerical usage is allowed now (sadly, including nfts)
|
|
#
?
Aug 4, 2022 00:30
|
|
|
precision posted:wait, it's okay to do that? At least with MidJourney you can. Ceramic Shot posted:I feel like the novelty value alone will kickstart its relevance in broader media. Definitely don't envy graphic designers now, but I'm kind of giddily anticipating what kinds of applications people way smarter than me will figure out. And honestly, it's the wild west out there, you can probably do whatever you like before the law catches up with the ramifications of A.I image generation.
|
|
#
?
Aug 4, 2022 00:31
|
|
|
precision posted:wait, it's okay to do that? dall-e allows commercial use but they demand you clearly label that it was generated with their stuff midjourney has special demands if you make over a million a year with your AI art however all those restrictions are probably unenforceable outside of banning you from the service because AI generated art can't be copyrighted (modified versions of AI art can be copyrighted though, so assembling them together into a comic makes the comic a copyrightable work)
|
|
#
?
Aug 4, 2022 00:32
|
|
|
"cover art for a vintage comic book from 1969. it is about Vixy, a fox-themed superhero. She has a sidekick who is dressed like a bat. They are posing on the roof of a skyscraper at dusk." the others weren't great, but i think we're on to something here! how would you get it to keep using the same style and/or characters?
|
|
#
?
Aug 4, 2022 00:32
|
|
|
precision posted:wait, it's okay to do that? Already answered but here is proof: 
|
|
#
?
Aug 4, 2022 00:33
|
|
|
i think i'm starting to be properly terrified of how good these things are. like, this is freakin nuts!!!!
|
|
#
?
Aug 4, 2022 00:34
|
|
|
precision posted:how would you get it to keep using the same style and/or characters? I think if you want everything to match well you'll have to experiment with finding a style and character description that is highly replicable in dall-e even then it's probably gonna be a struggle, probably need to do a lot of erasing bits and regenerating until you get exactly what you want if you're willing to bust out photoshop and chop things together that would help also doing a black and white comic would make it much easier to keep things consistent frame to frame
|
|
#
?
Aug 4, 2022 00:37
|
|
|
i wonder if the price of credits will only go up, or if they'll start offering an Unlimited package. $30 gets you 230 credits, which seems like more than enough to generate a pretty good size issue of a comic
|
|
#
?
Aug 4, 2022 00:37
|
|
|
precision posted:i think i'm starting to be properly terrified of how good these things are. like, this is freakin nuts!!!! And this is just the start.
|
|
#
?
Aug 4, 2022 00:38
|
|
|
precision posted:i wonder if the price of credits will only go up, or if they'll start offering an Unlimited package. $30 gets you 230 credits, which seems like more than enough to generate a pretty good size issue of a comic you might be surprised how quickly you drain those credits when you're actually trying to make something specific
|
|
#
?
Aug 4, 2022 00:38
|
|
|
precision posted:i wonder if the price of credits will only go up, or if they'll start offering an Unlimited package. $30 gets you 230 credits, which seems like more than enough to generate a pretty good size issue of a comic I honestly expect (as has already been mentioned earlier in the thread) that Dall-E proper is going to quickly be superseded here by other less-restrictive platforms. Now the novelty is starting to give way to practical use its just a matter of time. Dall-E's credit structure is dumb as hell as well, not helping things.
|
|
#
?
Aug 4, 2022 00:39
|
|
|
A hacky way to try to keep the style is to leave an unimportant corner from the original image and do inpainting
|
|
#
?
Aug 4, 2022 00:41
|
|
|
precision posted:"cover art for a vintage comic book from 1969. it is about Vixy, a fox-themed superhero. She has a sidekick who is dressed like a bat. They are posing on the roof of a skyscraper at dusk."
|
|
#
?
Aug 4, 2022 00:42
|
|
|
precision posted:how would you get it to keep using the same style and/or characters? Style is fairly easy but with both comics made by Kholsa, They Will Inherit Hell and The Party And Its Guests, they're short mood pieces. There's very little action or continuity between panels. Montage and voice over text do most of the heavy lifting to create meaning.
|
|
#
?
Aug 4, 2022 00:42
|
|
|
The creator of midjourney had an AMA discussion on discord and there was a user who is a film director and simply couldn't contain his excitement about the fact that he doesn't need concept artists anymore and that he doesn't have to deal with the ineptitude of a human artist anymore. Someone in the chat commented that AI should soon replace directors too.
|
|
#
?
Aug 4, 2022 00:44
|
|
|
KinkyJohn posted:The creator of midjourney had an AMA discussion on discord and there was a user who is a film director and simply couldn't contain his excitement about the fact that he doesn't need concept artists anymore and that he doesn't have to deal with the ineptitude of a human artist anymore. They will replace directors.
|
|
#
?
Aug 4, 2022 00:45
|
|
|
Trying some stuff   Here I tried something mentioned in a brief article on the linked DALL E dictionary site. Take a prompt:  Run it through OpenAI text playground by writing something like this. In their example they had the text writer describe A Starry Night in detail and used that.  Then copy the whole green part in with a small adjustment, and you get:   Well, I still think its worth a shot if you're trying something less complicated or want to really try and ape some artist's work
|
|
#
?
Aug 4, 2022 01:27
|
|
|
Now I have to plan how to spend all that money that will roll in. 
|
|
#
?
Aug 4, 2022 02:48
|
|
|
fez_machine posted:Abhay Khosla of Abhay Khosla's Bram Stoker's Dracula has been making comics using MidJourney and the results are pretty good. this is sick as hell, thanks
|
|
#
?
Aug 4, 2022 03:06
|
|
|
PITY BONER posted:Am I the only one that sees that its source for the fox man is distinctly from furry art? accidentally so, but probably unavoidable. search for any animal + adjective / verb / noun combo in GIS and you get furry poo poo mixed in only a matter of time before a focused effort produces an image generator trained exclusively on that diet. DallE 2 junk:  "laughing shrimp japanese restaurant logo"  "rolling soundwave, art deco, gold on black"  I forget the prompt for this one, but if this was a PS1 game cover I would have been all over it.  variation from uploaded photo of a handpainted logo resembling Porky Pig's retarded cousin, on the back of a food truck. The variation is deliciously demented looking.  "a beaver gas station logo"  "colorful cassette tape, 1980s poster art"  "modular synthesizer for cats" frumpykvetchbot fucked around with this message at 03:44 on Aug 4, 2022 |
|
#
?
Aug 4, 2022 03:14
|
|
|
i'm pretty curious what kind of legal fuckery is going to be attempted once the first 'free' system starts producing incriminating photos, let alone full forms of entertainment i could easily see it getting regulated with the intensity of say, CSAM, as soon as it becomes even a blip on the elderly political radars
|
|
#
?
Aug 4, 2022 03:18
|
|
|
AARD VARKMAN posted:i'm pretty curious what kind of legal fuckery is going to be attempted once the first 'free' system starts producing incriminating photos, let alone full forms of entertainment finally, the piss tape will be real
|
|
#
?
Aug 4, 2022 03:19
|
|
|
precision posted:finally, the piss tape will be real it was always real (NSFW) https://imgur.com/hWbkLqU
|
|
#
?
Aug 4, 2022 03:25
|
|
|
ymgve posted:it was always real (NSFW) lmao why is he just checking his phone and not like jerkin' it or whatever it's like, there is no part of this man that is a normal human being
|
|
#
?
Aug 4, 2022 03:28
|
|
|
AARD VARKMAN posted:i'm pretty curious what kind of legal fuckery is going to be attempted once the first 'free' system starts producing incriminating photos, let alone full forms of entertainment The worst unanticipated outcome is went it fully gets into the hands of the feel good fakers, god botherers, and alternative reality type folks peddling healing products, they already have a cavalier attitude to harm and faking poo poo. It'll just get 1000 times worse
|
|
#
?
Aug 4, 2022 03:31
|
|
|
what kind of words are people using to get like... the grody realistic horror watercolor type pics? i've tried all kinds of stuff like "a rusted cathedral in the background, a desolate cemetary in the foreground, a sad wizard walks along with his pet bat, gothic style" or "bloodborne style" and apparently it thinks "gothic style" is basically like, "bauhaus", and i don't mean the band
|
|
#
?
Aug 4, 2022 03:38
|
|
|
precision posted:what kind of words are people using to get like... the grody realistic horror watercolor type pics? i've tried all kinds of stuff like "a rusted cathedral in the background, a desolate cemetary in the foreground, a sad wizard walks along with his pet bat, gothic style" or "bloodborne style" and apparently it thinks "gothic style" is basically like, "bauhaus", and i don't mean the band 
|
|
#
?
Aug 4, 2022 03:44
|
|
|
weird, it didn't give me anything nearly that good
|
|
#
?
Aug 4, 2022 04:36
|
|
|
You can also try adding “by beksinski” to your prompts. He worked in oils, but I get the feeling that you might like the look of the results.
|
|
#
?
Aug 4, 2022 04:39
|
|
|
precision posted:weird, it didn't give me anything nearly that good here's the prompt quote:the grody realistic horror watercolor type pics? i've tried all kinds of stuff like "a rusted cathedral in the background, a desolate cemetary in the foreground, a sad wizard walks along with his pet bat, gothic style" or "bloodborne style" sometimes doing stupid poo poo like this works
|
|
#
?
Aug 4, 2022 04:42
|
|
|
Make sure you read the user manual as well. https://midjourney.gitbook.io/docs/user-manual#stylize-values There are a lot of hidden modifiers you can append to your text. You can increase or decrease how closely the generator will stick to your prompt, or let it go off the rails and do crazy stuff. You can feed it an image for it to take inspiration from when it generates your prompt. You can weight different words of your prompt with varying levels of importance, so if you go "purple dog with eagle claws", and you really want those claws in there, you can weight the words eagle and claws with more weight than the other terms.
|
|
#
?
Aug 4, 2022 04:54
|
|
|
precision posted:i wonder if the price of credits will only go up, or if they'll start offering an Unlimited package. $30 gets you 230 credits, which seems like more than enough to generate a pretty good size issue of a comic I guess you're talking about DALL-E, but Midjourney is giving unlimited credits for $30/month. I quickly caved and got that package; now I'm dizzily alternating between learning different tags like Bloody Hedgehog mentioned one post above me, and just experimenting.
|
|
#
?
Aug 4, 2022 05:32
|
|
|
Ceramic Shot posted:I guess you're talking about DALL-E, but Midjourney is giving unlimited credits for $30/month. I quickly caved and got that package; now I'm dizzily alternating between learning different tags like Bloody Hedgehog mentioned one post above me, and just experimenting. How long does MJ take to generate its images?
|
|
#
?
Aug 4, 2022 05:38
|
|
|
Rinkles posted:How long does MJ take to generate its images? About 30 seconds or so to produce a set of four images based on your prompt. Then you can (1) choose to individually "upscale" individual pictures to be more detailed, (2) ask it to "re-roll" and produce four more images based on your original prompt, or (3) produce four more images based on one of the originally produced images (this is very fast, maybe 15 seconds). Upscaling further takes maybe 90 seconds, and then you can upscale it a final level that takes maybe 2.5 minutes. It's loving crazy. edit. I timed a medium complexity prompt and it went like: Initial set of four based on prompt: 0:45 Upscaling one of the pictures 1:42 Re-rolling the initial four pictures: 0:45 (same as #1) Asking for four new variations on one of the pictures from the initial set: 0:17 Maximum upscaling: 1:39 (this was odd, usually max level upscaling takes significantly longer than initial upscaling. Probably a fluke.) Here were the two most interesting products:   The prompt was something like: a menacing roger rabbit bloodborne hunter with a saw cleaver, bloodborne style, old yarhnam, gibbous moon, hyper realistic, cinematic composition, cinematic lighting. Ceramic Shot fucked around with this message at 06:04 on Aug 4, 2022 |
|
#
?
Aug 4, 2022 05:41
|
|
|
precision posted:how would you get it to keep using the same style and/or characters? I would not be at all surprised if some time within a year or two there will be a client-side training option so you can feed AIs reference images of your characters. It's going to be necessary if folks are going to use them to create graphic novels and such.
|
|
#
?
Aug 4, 2022 05:51
|
|
|
Ceramic Shot posted:About 30 seconds or so to produce a set of four images based on your prompt. Then you can (1) choose to individually "upscale" individual pictures to be more detailed, (2) ask it to "re-roll" and produce four more images based on your original prompt, or (3) produce four more images based on one of the originally produced images (this is very fast, maybe 15 seconds). From an end-user perspective that sounds much better than DALL-E, which is great but sometimes feels like a slot machine with its output.
|
|
#
?
Aug 4, 2022 05:53
|
|
|
ymgve posted:it was always real (NSFW) precision posted:lmao why is he just checking his phone and not like jerkin' it or whatever unfortunately that is just a very well done fake, took forever to surface the source despite it being part of a segment on a BBC 3 show go to 28:45 to see the making of https://thoughtmaybe.com/blindboy-undestroys-the-world/ they did a great job, there was a journalist who tried to debunk it at one point and all their research did was make it seem even more real
|
|
#
?
Aug 4, 2022 06:13
|
|
|
|
| # ? May 29, 2024 19:32 |
|
|
note that midjourney's relax mode prioritizes people who have used it less, so if you use it a lot, you drop in the queue over the course of a month as and it gradually gets slower
|
|
#
?
Aug 4, 2022 06:18
|
|