|
deep dish peat moss posted:Midjourney has great respect for pixel art bears What is this 1, 2, 3, 4 thing??
|
#
?
Apr 4, 2023 23:25
|
|
|
|
| # ? May 28, 2024 13:10 |
|
|
Doctor Zero posted:What is this 1, 2, 3, 4 thing?? Midjourney Img2Txt  nonconformist art brut Negative prompt: amenable, compliant, docile, law-abiding, lawful, legal, legitimate, obedient, orderly, submissive, tractable Steps: 32, Sampler: DPM++ 2M Karras, CFG scale: 11, Seed: 520244594, Size: 512x512, Model hash: 99fd5c4b6f, Model: seekArtMEGA_mega20
|
|
#
?
Apr 4, 2023 23:30
|
|
|
PlaysGamesWrong posted:Holy poo poo, I'm keeping those for later tabletop use Here's a template for the prompt (MJv5): a [class] [animal] in [garmets] and holding an olive stick, in the style of dark [color] and light [color], neogeo, terrorwave, i can't believe how beautiful this is, full body, grid-based, toonami a warrior lion in armor and holding an olive stick, in the style of dark yellow and light blue, neogeo, terrorwave, i can't believe how beautiful this is, full body, grid-based, toonami  YMMV when experimenting with removing "and holding an olive stick", or with changing the [animal] to a fantasy race like elf or dwarf. Sometimes it works, sometimes it doesn't. It works much better with animals. a wizard elf in robes and holding an olive stick, in the style of dark yellow and light blue, neogeo, terrorwave, i can't believe how beautiful this is, full body, grid-based, toonami  Classes you wouldn't expect to wear "armor" seem to have a high chance of generating pixel art when "in armor": a rogue elf in armor and holding an olive stick, in the style of dark yellow and light blue, neogeo, terrorwave, i can't believe how beautiful this is, full body, grid-based, toonami  I think the style is mostly coming from Neogeo + Toonami (possibly terrorwave?) and the rest of it is influencing the composition and content, feel free to mix and match. deep dish peat moss fucked around with this message at 23:56 on Apr 4, 2023 |
|
#
?
Apr 4, 2023 23:45
|
|
|
https://onceuponanalgorithm.org/con...-_d0O_H4h8qceZo I've got this project I'm working on which would be like a $200 million live action movie. But I've looked at some critically acclaimed anime like Cyberpunk Edgerunners, and I may be missing a lot here but the majority of what I see doesn't look more technically complex than a foamy the squirrel flash cartoon from 2005. Low framerate, lots of static scenes where only a mouth moves, or one person walks past static backgrounds... Is SD good enough right now that I, an artist or two, and some voice actors could make a respectable anime with some work? I've seen the Corridor Crew AI anime, but they're a pro FX crew and their video is annoying so I didn't watch it all. The only other stuff I've seen around is very crude quick test clips. Here's my AI project shot vs AI anime version. MJ Niji 5 is kinda nice.  
|
|
#
?
Apr 5, 2023 01:20
|
|
|
Roman posted:https://onceuponanalgorithm.org/con...-_d0O_H4h8qceZo No, you are very much wrong. Like there are times when even highly quality anime does some cut-costing for animation, but that's just to preserve their their budgets and time for the really amazing animation. Anime that just uses those cut cutting measures are usually considered very bland.
|
|
#
?
Apr 5, 2023 01:35
|
|
|
IShallRiseAgain posted:No, you are very much wrong. Like there are times when even highly quality anime does some cut-costing for animation, but that's just to preserve their their budgets and time for the really amazing animation. Anime that just uses those cut cutting measures are usually considered very bland.
|
|
#
?
Apr 5, 2023 01:43
|
|
|
Roman posted:Is SD good enough right now that I, an artist or two, and some voice actors could make a respectable anime A thorough complete and resounding YES quote:with some work? HAHAHAHAHAHAHAHAHA
|
|
#
?
Apr 5, 2023 01:46
|
|
|
KwegiboHB posted:Midjourney Img2Txt nonconformist art brutalize lawfully telling your bros they look submissive and tractable
|
|
#
?
Apr 5, 2023 01:50
|
|
|
Roman posted:https://onceuponanalgorithm.org/con...-_d0O_H4h8qceZo Duck the haters. Do it. You know how South Park started out.
|
|
#
?
Apr 5, 2023 01:51
|
|
|
Roman posted:https://onceuponanalgorithm.org/con...-_d0O_H4h8qceZo There's a lot to animation, but one of the key things every good animator learns early on is that complexity, technical detail, and realism are (generally speaking) not good for animation. It has been pulled off well (shining example: rotoscoping in things like A Scanner Darkly), but it's often the exaggerated lack of realism that makes animation feel colorful and alive. AI is a long way off from matching the best examples of human animation, but you're not wrong that the vast majority of animation is just not that good - chiefly due to how expensive it is to produce. I don't work with SD enough to have a valid opinion on it but I don't think it's good enough out of the box that you could make respectable anime, but look at what posters like e.g. TIP are doing ITT to see what's possible with SD animation. deep dish peat moss fucked around with this message at 01:54 on Apr 5, 2023 |
|
#
?
Apr 5, 2023 01:51
|
|
|
KwegiboHB posted:HAHAHAHAHAHAHAHAHA deep dish peat moss posted:but it's often the exaggerated lack of realism that makes animation feel colorful and alive. AI is a long way off from matching the best examples of human animation, but you're not wrong that the vast majority of animation is just not that good - chiefly due to how expensive it is to produce. Roman fucked around with this message at 01:55 on Apr 5, 2023 |
|
#
?
Apr 5, 2023 01:53
|
|
|
TIP has been working with a custom text2animation model ITT, reading their posts may give a better understanding of the kind of effort you're looking at - here's the most recent:TIP posted: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. Pure speculation from me but: It'll probably be another year or two (assuming no major roadblocks in AI animation show up) before AI will be able to realistically do a passable version of the kind of animation you're looking for. A "more realistic" way to do it (right now) would be hand-drawing an outline for each frame and then using SD's capability to use that to guide the shape of the generation to fill it in (I forget the name of that tool, not a SD person myself). That's realistically "small team" work but an enormous amount of work hours for the individuals (sketching and generating each frame individually, stitching them together into an animation) edit: But also as an incredibly sincere suggestion for small-team low-effort animation, consider how Squigglevision works The only thing you need for something to 'feel' animated is kinetic movement, even if there's no actual animation. You can easily turn a series of still images into something that feels animated, without actual animation. This type of animation (not Squigglevision specifically, but "mostly still images with kinetic movement") worked well for tons of Adult Swim shows, video games (think Darkest Dungeon), digital versions of comic books, etc. In game development it's generally called "Juice" and you can find plenty of videos/articles about how to add "juice" to your game that could give you some ideas which could be translated to animation. So don't let the accessibility of Animation turn you away from following your creative dreams  You could realistically do something with this concept using AI and a little bit of Human elbow grease right now.
deep dish peat moss fucked around with this message at 02:12 on Apr 5, 2023 |
|
#
?
Apr 5, 2023 01:57
|
|
|
deep dish peat moss posted:The only thing you need for something to 'feel' animated is kinetic movement, even if there's no actual animation. You can easily turn a series of still images into something that feels animated, without actual animation. This type of animation (not Squigglevision specifically, but "mostly still images with kinetic movement") worked well for tons of Adult Swim shows, video games (think Darkest Dungeon), digital versions of comic books, etc. In game development it's generally called "Juice" and you can find plenty of videos/articles about how to add "juice" to your game that could give you some ideas which could be translated to animation. So don't let the accessibility of Animation turn you away from following your creative dreams 10 year old classic of indie game development https://www.youtube.com/watch?v=Fy0aCDmgnxg
|
|
#
?
Apr 5, 2023 02:16
|
|
|
deep dish peat moss posted:A "more realistic" way to do it (right now) would be hand-drawing an outline for each frame and then using SD's capability to use that to guide the shape of the generation to fill it in (I forget the name of that tool, not a SD person myself). That's realistically "small team" work but an enormous amount of work hours for the individuals (sketching and generating each frame individually, stitching them together into an animation) But even with the figure posing tools, there's the question of lip-sync and facial expressions. The only reason I'd want to do animation is if those things could be done well. Here's my weirdo cult characters with the power of anime.
|
|
#
?
Apr 5, 2023 02:18
|
|
|
Roman posted:https://onceuponanalgorithm.org/con...-_d0O_H4h8qceZo Did you Corridor's explainer video where they gave a breakdown of how they made Rock Paper Scissors? https://www.youtube.com/watch?v=_9LX9HSQkWo
|
|
#
?
Apr 5, 2023 02:19
|
|
|
Roman posted:Is SD good enough right now that I, an artist or two, and some voice actors could make a respectable anime with some work? You have, as another option for AI generated animation, Deforum. I posted a couple I've done earlier in the thread. Though what you'll get is going to be very different stylistically from any traditional animation. Mine are all with high denoising and adjusting the image every single frame because I was going for that radically different style but, with the opposite, and especially with using a mask video you can get more consistency. There are a lot of very bad "tech demo" Deforum type projects out there too, but with it the skill ceiling is very high so I wouldn't completely base its capabilities off of those.
|
|
#
?
Apr 5, 2023 02:36
|
|
|
I wouldn't say so far reverse engineering prompts from similar photos is a more accurate method than just describing what you want on your own terms Describe has however taught me midjourney loves to think in terms of "core" smilecore, nerdcore, hispanicore etc
|
|
#
?
Apr 5, 2023 03:31
|
|
|
deep dish peat moss posted:TIP has been working with a custom text2animation model ITT, reading their posts may give a better understanding of the kind of effort you're looking at - here's the most recent: There's also Joel Haver's method he's been using for a couple years now: https://www.youtube.com/watch?v=tq_KOmXyVDo Basically he rotoscopes keyframes of animation on top of live action, and then uses basically style transfer from the software to apply those hand-drawn frames of art onto the subsequent frames of video. It tends to break when there's too much movement, though, but that makes it perfect for lip-syncing. Combining that with SD or MJ keyframes could yield very interesting (and cost-effective) results. That's a potential way to animate small details, if you can find Andy Serkises rather than just VAs.
|
|
#
?
Apr 5, 2023 03:59
|
|
|
I just wrote three useful tools in python to help automate my animation workout, but I don't know python  I'd heard ChatGPT can code, and it's true! However it is just as limited as I expected. If you ask for something really simple it can give you something that works, but it really starts to lose the train of thought quickly if you ask for something more complicated. It will also just make up fake APIs and poo poo if you ask it to do something that's out of its training data. It seems the best way to go is to ask it do a very basic version of what you want, and then once you've tested it and it works open a new chatgpt and describe the changes you'd like made to the code and then paste the code in below that. If you just ask the first chatgpt to modify the code it wrote it seems much more likely to go off the rails. Even the code that works can be pretty dumb, like in the video processing code it just filled up the memory with old frame data until it died. So knowing some programming is still useful, I told it to make the code memory safe and it fixed that. Anyone here subscribe to open AI for the better chatgpt and gpt4? I'm wondering how improved they are for programming, if they're much better it might actually be worth the money.
|
|
#
?
Apr 5, 2023 05:48
|
|
|
I've started messing around with the prompt matrix in Stable Diffusion, it's pretty fun. This is "retro poster for a 1970s car doing stunts while driving|on mars|in a dome|in a cyberpunk landscape|in hell|in front of a diner|in an apocalyptic wasteland" Linked because it's 4k pixels wide https://i.imgur.com/aYf81z7.jpg
|
|
#
?
Apr 5, 2023 06:25
|
|
|
Here's a good prompt I discovered: "A Trump urinal, mouth open wide"
|
|
#
?
Apr 5, 2023 12:20
|
|
|
This is going to sound insane but it makes sense if you know my group of friends. And he's totally cool with this and thinks it's funny as hell. The one friend who gave me just complete permission to use him for training experiments and stuff (hes got a poo poo ton of training data because he hosts some of these documentaries), it's his birthday pretty soon. So, I'm making him a book, I'm gonna get it all professionally printed. And the book is going to be entirely AI generated. So I ripped several thousand erotic fanfics from ao3, and then wrote a thing that replaces all the names with his name, "Cody." Then I trained GPT-Neo on it all. It kinda hosed up and replaced some non-names with Cody too so the model just invents new words that are variations of his name but that just adds more flavor. It made a song (NSFW)  And the non-song stuff that's just like erotic fanfics (VERY NSFW) 
|
|
#
?
Apr 5, 2023 12:26
|
|
|
So Facebook released some Segment anything models https://twitter.com/MetaAI/status/1643599800414380038 https://github.com/facebookresearch/segment-anything. Its very cool, and lets you create a mask for all objects in an image or select specific objects to mask. I'm sure somebody will make an automatic111 extension to make in-painting easier too.
|
|
#
?
Apr 5, 2023 15:46
|
|
|
where's the main ai thread or chatgpt thread?
|
|
#
?
Apr 5, 2023 18:04
|
|
|
Mescal posted:where's the main ai thread or chatgpt thread? I think it's kinda just this one. There's been plenty of chatGPT discussion in here.
|
|
#
?
Apr 5, 2023 18:05
|
|
|
We talk about making computers do things normally thought of as human. Images, speaking, text, I'm sure we'll talk about making text to smell at some point.
|
|
#
?
Apr 5, 2023 18:11
|
|
|
KinkyJohn posted:Here's a good prompt I discovered:      
|
|
#
?
Apr 5, 2023 18:23
|
|
|
Oh it's Your Pretty Face is Going to Hell! If anyone has a workflow for video to video here's a clip well the start of it is it goes on but the only just this scene was on Adult Swim directly which is probably going to be harder to rip than just ripping and trimming. Probably the behind Satan shot is best? At least from season 1, they do this bit again. https://www.youtube.com/watch?v=wZBMvAMhIZs I feel like I'm asking a lot but having seen what some people have been pulling off with guiding video maybe I'm not?
|
|
#
?
Apr 5, 2023 18:32
|
|
|
StarkRavingMad posted:I think it's kinda just this one. There's been plenty of chatGPT discussion in here. can't believe that. Chatgpt thread
|
|
#
?
Apr 5, 2023 19:11
|
|
|
Mescal posted:can't believe that. Chatgpt thread We don't really need two threads lol.
|
|
#
?
Apr 5, 2023 19:30
|
|
|
IShallRiseAgain posted:We don't really need two threads lol. Three, there's a separate thread for discussing the ethics of AI. Also TG has its own AI thread specifically devoted to only Trad Game nerding out.
|
|
#
?
Apr 5, 2023 19:39
|
|
|
And most goon discords also have an AI dumping channel to just post dumb poo poo you make that's not fully worth posting.
|
|
#
?
Apr 5, 2023 20:28
|
|
|
pixaal posted:…dumb poo poo you make that's not fully worth posting. This is GBS. Thats not a thing.
|
|
#
?
Apr 5, 2023 20:58
|
|
|
atomicpile posted:This is GBS. Thats not a thing. Strange prompt, has some early '90s alternative music album vibes to it.  edit: drat second run was something that is worth posting, the AI is dumb and ignored the prompt. 
|
|
#
?
Apr 5, 2023 21:04
|
|
|
atomicpile posted:This is GBS. Thats not a thing.  Why do you want to see poor quality Steve Brule: Hunk Wrestler? Mike Tyson boxing a room of angry hornets?  yeah it had no idea how to combine these concepts A for effort AI
|
|
#
?
Apr 5, 2023 21:05
|
|
|
Fuzz posted:Three, there's a separate thread for discussing the ethics of AI. Also the thread for discussing AI generated channels on Twitch
|
|
#
?
Apr 5, 2023 21:51
|
|
|
Have a link for that?
|
|
#
?
Apr 5, 2023 21:57
|
|
|
Just a warning, the TG thread is full of AI haters and the mod there doesn't give a poo poo if they just bully people, because they too are an AI hating bully.
|
|
#
?
Apr 5, 2023 22:00
|
|
|
oh no they posted posts  i hope you got out unharmed e: lol i read the thread and it was like half a page where two posters disagreed with you RPATDO_LAMD fucked around with this message at 23:02 on Apr 5, 2023 |
|
#
?
Apr 5, 2023 22:32
|
|
|
|
| # ? May 28, 2024 13:10 |
|
|
RPATDO_LAMD posted:oh no they posted posts Kindly eat your own head. Seriously though, someone posts that they want to make art for themselves, people show up and scream at them and call them a thief for having the nerve to be interested in AI in the AI thread, and the mod dgaf calls the overt harassment for discussing the subject of the thread "legitimate discourse" and wanders off. There's a reason that thread is on like page 3. The OP is Perma banned and you literally can't discuss the topic without idiot grognards making posts like this.
|
|
#
?
Apr 5, 2023 22:40
|
|