Using Krita AI Diffusion as an artist

Plugin: github.com/Acly/krita-ai-diffusion
Krita: krita.org/en/download/
Animesharp Upscaler: openmodeldb.info/models/4x-AnimeSharp
All my old stuff: patreon.com/posts/all-my-old-stuff-100466287
My old Stable Diffusion guide: patreon.com/posts/using-stable-as-100514304
Previous video: • Using Stable Diffusion...
Stable Diffusion glossary: stable-diffusion-art.com/glossary/
Voice(s) (schizo poorly audio mastered Discord/Xbox Live party): Bark
github.com/rsxdalv/tts-generation-webui

[@KimForsberg]

That's a really smart way to basically get a virtual assistant to support in detailing your roughs so you can spend your time where it matters, composition, design, and concept, and in the end, support in cleaning up the final result after manual detailing! All while keeping to your own style.

[@penwrythe]

That's interesting! To me, as long as Ai is used as a tool, not a replacement of artists, I'm fine with it. My style is developed already, but some like this would be useful if I need to make a quick mockup of an idea, then fully draw it out later.

[@TJPet2]

I think when it comes to this, it depends on how it's used person A : could use ai to do all the work for them. do a basic drawing, let ai do the rest, make your own fixes, and let ai finish up the drawing. i think doing this method might make the artist depended on it and their own skill might suffer from it person B : could do a full drawing, have ai try to improve it. , and instead of having it change your image, the artist do the changes themselves. basically, they are slowly refining their own skill using ai as an assist incoming with additional option of how to improve in their work. just thinking ai would more or less build or break a person depending on use :) OH, a side note, ever thought about using design doll or some 3D modeled pose software to create a pose, sketch over the pose, and have ai generate the image? :)

[@nodonmai]

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.

[@asciikat2571]

nice one !

[@apgamer4053]

Finally another person that dosent hate befor they try

[@lefourbe5596]

now that is what i've envision back when training went available in 2022. i'm no artists but i find enjoyement generating thing and as i learn new method and try new model i get closer and closer to acceptable result. the AI feel is a thing i want to avoid ultimately but the devil is in the details. i'm far from it and i'm not concerned. most of the time, there is manual work to put before and after the generation. consistency is key and coherent **interesting** animation cannot be fully automated. tweetos give too much intelligence to this raw algoritm. keep on drawing ! prompt is accesory.

[@8561]

Great tutorial!! Very clever to train your own custom LoRa.

[@scroptels]

really cool, would use it but i got no dedicated graphics card so i guess i'll stick to hand made anyway.

[@MythosMasonry-fq1up]

let's hope you don't get killed by the other artists

[@abcdef-vk7si]

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.

[@alinbot5836]

@@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.

[@digitalasylum369]

Great video!

[@AUselessSapphic]

Now thats a tool, not a cheat. I like this approach

[@sownheard]

that is great :D

[@wizards-themagicalconcert5048]

Great video and content mate ! Is there a place to get these Loras ? And,how can you train a Lora on your own style ?Maybe you answered this question already in another place,apologies ! Anyway,good stuff here,Subbed !

[@nodonmai]

Civitai is the current main place to get checkpoints and Loras. For Lora training, I recommend looking up tutorials here on KZbin, 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 ™

[@nodonmai]

kzbin.info/www/bejne/rpy2qGBjpZ2Xja8

[@skayakitty625]

fascinating. ai voice too yea? i won't bother with a discussion on my feelings abt ai art (broadly negative) but this is an excellent demonstration of the tech. thank you.

[@xenn2996]

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

[@nodonmai]

@@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

[@jackrabbit1704]

Thanks, AI is such an OP tool for artists. People can pointlessly hate AI all they want and be stuck using stones to build a house, good on them but I'll be using electric tools.

[@Thesamurai1999]

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.

[@jackrabbit1704]

@@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.

[@YouTubeDeletesMyCommentssss]

@@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.

[@jackrabbit1704]

@@KZbinDeletesMyCommentssss 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.

[@YouTubeDeletesMyCommentssss]

@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.

[@dadadies]

Thank you for the video. AI is great especially for real artist but not so great for 'capitalists (focused) artist'.

[@Kongongongg]

Based

[@urgyenrigdzin3775]

nicely put! 👍👍👍

[@niszolen5923]

?

[@alexisleskinen6090]

What is bro yapping about

[@Flazermov]

AI ARTIST IS THE FUTURE!

[@Thesamurai1999]

Not gonna get much better at drawing using methods like that. It just encourages quick lazy sketches for the AI to solve the rest of the problem. The AI literally doing all of the rendering and fixing most of the proportional mistakes. But a great method for corporations who are looking to pump out as much art as possible.

[@figuredrawing4912]

A.I. is the fast food version of REAL art.