Based

? I've been drawing all my life and even show my non AI stuff, try again.

Yeah, like the one I use in the video.

1:11

Using AI to make art for you isn't so different from paying artists to make art for you. If you are against both, then I can respect that madness. But if you are only against AI in this case, then... well...

it could be his voice you can clone your own voice using capcut.

@@xenn2996 Nah it's just one of the Bark male presets, but it's not as consistent as XTTS and every line read ended up somehow swapping characters, so it's really like 5 different voices

Civitai is the current main place to get checkpoints and Loras. For Lora training, I recommend looking up tutorials here on KZfaq, because even if it's relatively not that hard, it still can get really deep and technical. I went a bit over this in my PDF guide, but also I'm (very slowly) making a video about models that also covers this, soon ™

kzfaq.info/get/bejne/rtGDqpBe1Z7Ih4E.html

It's mocking those critiques specifically because they're dumb and unfounded. It's like mocking the idea that cameras and photographs work by stealing your soul, especially after explaining how they really work multiple times, and knowledge about their workings being readily available. Other arguments and concerns against AI are more nuanced and valid and worth arguing. It's the raw functionality of the tech that is more objective. I get people not understanding how it works, especially the math and code, and I don't get those either. I couldn't even begin to tell you how the Liquify tool works at a code level, but I know it objectively moves pixels around, it's like that for AI. I'm not a genius, in fact I'm very dumb because I enjoy fart jokes, but if even I could learn how it works by just using it and reading about it and watching other videos, then so can anyone. Hopefully the video itself still shows that the idea of it working by collaging and that there is no control is really dumb. Where are the images being pulled from as I'm manually drawing, especially if my Lora's dataset doesn't have anything close to it, let alone the base training data from LAION? That's the point of the video + some meme farts.

​@@nodonmaiI do agree with you that those arguments are invalid. What I don't agree with is the way it was expressed in the video. I mean, you know it's going to offend some people, don't you? (and then they won't watch the rest of the video, which is a shame). I think a more respectful approach could help foster better understanding and dialogue. We Stable Diffusion users don't need more hate from anti-AI people, nor giving them more reasons to hate us (even if the reasons are just farts ^^U).

@@juanjesusligero391 Oh yeah, it's absolutely on purpose. It's meant to mock the lowest of the low stubborn redditor, tumblr and twitter types. Anyone with at least 1 braincell should (hopefully) understand. For less memey videos there are the other two ones I made before this, and eventually another one I've been meaning to make.

​@@nodonmai :( I understand the frustration, but mocking others often backfires. It's like trying to put out a fire with more fire-it only makes things worse. When we ridicule people, we lose the chance to have a meaningful conversation and help them understand our perspective. A more respectful approach could help bridge the gap and foster better dialogue. I already watched your other videos, and I liked them a lot. Thank you for sharing them! :D

​​@@juanjesusligero391Can you detail that approach in full? Because last time I checked, the types of people that are against AI are religious about it to the point where you'll be burned at the stake and demonized for having a normal conversation about AI integration with art.

Yeah, over reliance of any tool can end up hurting more than helping. It’s all about balance and goals.

Like anything, skills only degrade if you let them, or if it's worth letting them degrade in favor of learning a better alternative. Digital art in general is like that, File > New Canvas in itself saves you from buying a canvas and priming it, let alone making it from scratch, and especially the latter involves skills that might get lost when moving fully to digital art only. With this method in the video, you control the back and forth between flesh and machine, and can also learn in various ways from it. With a Lora for example, AI can show you your "drawing quirks" either good or bad and you can decide which to keep or change, it can also fix your anatomy (to a point), composition, coloring and what not, and you can also learn by comparing what you do and what the AI does. Yeah you can use anything as a base image including 3D models, the workflow of base image > refine > manual tweaks > refine > manual tweaks >repeat is very simple and can apply to anything.

Sure, but don’t be surprised the day when no one will hire you or be amazed at what you do simply because even your grandma can do it at the click of a button.

@@Thesamurai1999 If a grandma can mass produce art faster than me then by all means, hire her over me. There's no shortage of art, imagine manga, anime, movies, video games, everything 10 times faster because of a tool. Just because bread began getting manufactured on an assembly line doesn't mean society suffered, that was a Neo-Luddite fallacy.

@@jackrabbit1704I am imagining it and it sounds all very uninspired, anything mass produced is proven to be crap and repetitive because it inevitably becomes about the money, the vast majority of AI “work “ I’ve seen is repetitive AI and very uncanny pictures of girls. And your right society didn’t suffer, the bread was just crap. Another question, what actual art will AI have when the internet is saturated with only images it’s generated? Also why are you using the word ART? when the term literally means inspired by HUMAN IMAGINATION. Everyone hyped by AI will be crying when they realise how the entertainment industry abuses it to the max just to push half assed content down everyone’s throats, if movies weren’t bad enough already.

@@KZfaqDeletesMyCommentssss 1. The device you're typing on to "explain" your fallacies was a tool to better society. It fast tracks the sending of messages, unless you think sending me stone tablets would be "better", then that's your opinion. 2. Show me the bread that isn't made with an oven, or even automatic ovens. Bread before tools was full of dirt or rock solid. Just because you can't find a company to use the proper tools and make good bread doesn't mean it doesn't exist. When I lived in Japan, the mass produced bread was amazing. 3. AI learns just like humans learn, when a human wants to draw picasso art style, they study picasso art style. When AI wants to draw uniquely in persona 5 game art style, it learns that style. AND can create unique things based on seeds, or an artists inputted portfolio. 4. That's the great thing about capitalism, people will spend their money on what's good, and what isn't wont sell, so they'll have to get better or go out of existence.

@jackrabbit1704 We've already seen what it looks like when companies just want a quick money grab, unfinished, broken, lazily written games, with no passion behind them, the same goes for movies, now imagine a world where all of that is extremely quickly rehashed, you'll have nothing unique, heartless passionless, stories, "art" games and movies, in which there was no blood, sweat or tears, yeah I'm somehow very unenthusiastic about a world saturated with uninspired crap where nothing is appreciated anymore because everything is fake. AI is completely dependent on a dataset. It rehashes solely based on what it sees, so it's not taking inspiration or coming up with new ideas. AI doesn't want to draw it draws what we want, which, for the most part, seems to be a bunch of waifus. Great fast tracks the sending of messages so we don't have to actually socialize face to face. Yeah, that's done us a world of good, Have you seen the world lately? Comparing phones and bread to art shows you have little to no value for actual art, which explains a heck of a lot.

It was a response to the equally childish "pick up a pencil" comments that are rampant on AI videos by those inclined to the Anti-AI way of thinking. Is it only okay for one side, are AI artists supposed to act perfectly? Man chose to show instead of talk.

@@pinip_f_werty1382 I understand but, two wrongs don't make one right. At least that's my opinion on the matter

I think it's a pretty awesome response 😂

For me this tool helps me to become an artist and understand how things work, because before that I would just draw scribbles with no result.

@@abcdef-vk7si Then I suggest to look at professional artists and observe and study their work. Ai usually never gets the anatomy right and there are some problem in understanding where the light comes from and shading so learning from professional artists would be much helpful and much better. there are also websites about anatomy and lighting made by people and they are looking much better than AI.

Based

nicely put! 👍👍👍

?

What is bro yapping about

Yeah character Loras help a lot with detail consistency. It's been a long while since I've tried animation, but doing quick sketches and then putting them through img2img/refine sort of worked when I experimented with it, and that was before Loras too. I also tried doing puppet-like animation using mostly just selection tools and liquify. That worked a bit better for consistency. Though I'd rather do that with proper puppet and skeleton animation programs, using AI to make or refine the base puppet, then manually cutting out the body parts, eyes, hair, etc. and using AI to clean the cut outs before moving to the animation program. The biggest issue for using img2img for animation at least for me, aside from detail coherency and consistency, is that colors tend to noticeably shift after too many img2img/refine passes and especially after multiple frames. So for maintaining 100% on model color consistency, it requires a bit more manual work.

@@nodonmai i was thinking of training a lora of a b/w manga character then creating 3d model of said character using ai then pose it to make 2.5d MMV videos do you think that's feasible ? , from my experience the 3d model doesn't have to look that good as inpainting and img2img can do wonders as long as the rough shape of the character is similar

@@thecolorowl5923 Yeah I think it's doable, it might be better to do with proper animation workflows and extensions either in ComfyUI or A1111, but it should still be feasible in Krita itself. The workflow of starting image (sketch, drawing, generation, 3d model, etc.) > refine > manual tweaks > refine > manual tweaks > repeat is very basic and technically it should work for anything, and you can use ControlNets in Krita AI Diffusion too. I've been meaning to get more into Stable Diffusion assisted animation but haven't yet, so I can't offer more specific animation advice at this moment, but there should be better AI assisted animation tutorials here on KZfaq. This channel seems like a decent starting point, it has some AI assisted animation with 3D models tutorials. www.youtube.com/@mickmumpitz/videos

I go over it a bit in the PDF guide, but I honestly recommend checking out Automatic 1111 and Krita AI Diffusion setup guides here on KZfaq, which can explain the technical aspects way better than me. It's not that hard aside from installing Git and Python, though as an artist I admit it can be a bit daunting at first. The first main roadblock for a lot of artists, aside from the more technical aspects, is having a decent Nvidia GPU in the first place, to even be able to use Stable Diffusion without even more complicated fiddling that AMD, Intel and Macs have to do. For models, especially as an artist starting with Stable Diffusion, I recommend a checkpoint and Loras that are already as close enough to your current art style, or close to the style of the art you want to make. You can find plenty of checkpoints on Civitai, it's just a matter of browsing until you find the one you like and downloading them. However, the best checkpoint for any artist is absolutely one that's been trained/finetuned with your own works. A Lora is faster to train than a checkpoint, and you can also merge it to a checkpoint for convenience, so I recommend training a Lora over a checkpoint. I sort of go over all these tech aspects in the PDF guide too, including plugins/extensions, but I still recommend checking out other more in depth tutorials here on KZfaq. I think I linked some of them in the guide, but there might be newer and better guides that are even easier to understand now. You can get the guide here in this link now without having to sign up for a Kofi account, it should be a direct download: patreon.com/posts/using-stable-as-100514304 Though currently for an artist I'd recommend Krita AI Diffusion over Automatic 1111, but both have their uses. Setup is even simpler as it's just unziping and copying over a couple of files. There's a showcase on this channel and also other setup tutorials on KZfaq. github.com/Acly/krita-ai-diffusion For overall learning about Stable Diffusion basics, there's also this site: stable-diffusion-art.com/glossary/