This One Simple Plugin Adds Realtime AI Assistance to Krita

in just two steps anyone can draw an owl step one draw some circles we got a circle and another Circle step two draw the rest of the funny owl yes hello and welcome to More nerdy rodent geekery where today I'm diving into the world of live stable diffusion yes you too can create masterpieces in almost real time thanks to the power of crater and comfy UI but look M there's absolutely no spaghetti this uses LCM so if you've already got comfy UI then you will be halfway there if on the other hand you don't even know what comfy UI is then don't worry as this will just install everything for you you get all the benefits of an app like creater such as all these different brushes and tools and everything you normally get with a painting package and almost anyone can scribble even a rodent moving and posing skeletons in real time is also another thing you can do and adds quite a new dimension to your image Generations it does so much more too but for this video I'm simply going to focus on what you need to get this up and running in order to generate images in real time but before we get too carried away with all the fun stuff you can do with it it is worth making sure you've got a computer which can run stable diffusion thankfully the requirements are a pretty meager 6 gig for vram and with regards to OS either Linux or Microsoft Windows is a good choice with experimental support for mac o also being available you'll need Creer installed for this which you can do via a single click in the free software store on Linux or you could do it via the Creer website as well well links are down in the video description 5.2.1 is the current release which is also the version they recommend I'll give you a second to do that now if you need one all done excellent with Creer installed and running you'll need to check one bit of information before you add in this new extension so head on over to settings and then the first option there configure creater over on this right hand side you got the resources Tab and that resources folder the whole path there is the thing that you need because that is where you're going to unzip the plugin the plug-in you can download from the GitHub page there it is download the plugin like it says unpack into your py Creer directory which is this one we just saw here the resources folder once you've done that it should look something like this so there is my resources folder and it's got the AI image diffusion file that's the configuration file and also that directory there aior diffusion with all the files in the next step is to enable the plug-in and this involves a few restarts of Crea so go ahead and restart again now that you have unzipped that archive and then head back into settings again and configure Creer but this time you're going to scr right down to the bottom where it says python plugin manager now yours will be unticked so there AI image diffusion so what you want to do is tick that and then click okay and restart creater once again we'll start with a new image at 512 by 640 as that is a reasonable size for both stable diffusion 1.5 and for image generation speed okay now you're going to need to have this show up first of all and that's going over to settings make sure you got show Dockers ticked and then this Dockers menu AI image generation the one at the top there if you tick that then your Docker will appear now as you can see there it's saying connection attempt failed click below to configure and reconnect which is exactly what you need to do whether you've got stable diffusion installed already or not here under connections it'll be on local server managed by the creater plug-in by default and these are the two options you can either have it manage it or you can manage your own local server if you're going to go with the local server managed by the plug-in this is the easiest option and the best one if you don't already have comfy UI installed because this will download everything for you it's about 10 gig in total as you can see you got lots of options you can tick stable with effusion 1.5 x sell lots of models which overall is about 10 gig all you have to do then is Click install wait for it to install and you're done you can move on to the next video Chapter which deals with plug-in options if you're already very comfy like me I am a very comfy nerd and that means you'll already have a bunch of models and stuff and you don't want that option so choose this other option connect to external server local or remote now you can see we've got a problem there error connection refused that's because I haven't got comfy UI set up yet back over on their GitHub page under the optional custom comfy UI server you'll see this link here please check the list of required extensions and models for this version which is 1.7.1 at the moment there are four custom nodes you will need they're listed under the required custom nodes section you'll need some control net pre-processors IP adapter ultimate SD upscale and external tooling nodes as well if you've used any of my free comfy UI workflows from the a very comfy nerd website then you'll likely have three of these installed already anyway whatever the case the easiest way is to install these using comy UI manager it's got a search facility so there I'm searching for the external tooling obviously I've already already got it installed but that way you can install all of the extra nodes along with those custom nodes you'll also need the models they list as required I already had everything apart from the specific upscaler they use and likely you will too as it's all pretty standard stuff now one thing to note here is how it handles model names and directories because some things are fixed such as these LCM luras while others use a substring search to match part of the default file name this means it should find all of your do bin or do safe tensor files as long as the file names don't differ too much if you are having issues with it not finding any models then the troubleshooting section at the bottom there is available for help like it says open up the client. log file to get more information about what is happen happening because the log file is in a directory which starts with a DOT you got to make sure you have show hidden files enabled otherwise it will look like it isn't there so there it is do logs and you got the client dolog and server.log in there here is my client. log file and as you can see it's nicely listed all the files it found for the control Nets IP adapter and LCM if you have any errors in here those should indicate what you need to fix okay with all the nodes installed and their models in their rightful places ensure comfy UI is up and running then when you pop back into here back into your server configuration and click connect this time it should go green and say connected leaving just a few other bits to explore other than your connection options there are a few other things available in the St menu this is where you can do things like change the model lauras prompts and vae all very standard stuff if you're used to stable diffusion click these tiny little arrows down here and you'll see even more settings personally I'm absolutely fine with the default but if you like to Tinker that is where you can diffusion interface and performance I've also left on the defaults with diffusion having some selection options interface has some prompt settings and performance has some options for history and also for different devices as mentioned I basically just leave those all in the default automatic is absolutely fine with the plug-in installed comfy UI running and all of your options set it's time to get drawing for live mode you can click the tiny little box up there and you'll see the option for live live you've also got a strength bar the D noising strength there by default at 30 and a seed which you can randomize by rolling the dice the cursor is blinking there ready for me to enter a prompt so let's enter one which is awesome now prompt in place we can just click the play button and we're ready to rock and roll Marvel at my art skills as I deafly wield my digital brushes in order to create create a masterpiece okay so are you ready first of all let's make this brush a little bit smaller let's go down to six and uh yeah let's see if we draw something like this and then uh maybe if I give him a little hat there we go now as we know rodents have nearly perfectly triangular faces there we go now this isn't looking very much like a rodent gentleman yet is it well this is where the strength bar comes in if we go all the way up to one obviously it's going to change the image completely so I like it a little bit lower typically you know somewhere around 50 60 70% is pretty nice around there it's sort of got a a bit of a sort of cartoony sketchy feel to it and the higher up it goes more photographic and realistic so that's pretty cool put the little legs on have him waving his arms about there we go all far too much fun maybe if I color the Hat in it's going to give him that hat oh yeah that's more like it that is more like it see how smart that hat is now and if you think that's cool then it gets even cooler because if you remember you can use all of those control Nets in real time as well just as an example if you add a new vector layer there's a tiny Arrow there add vector layer then we'll add a new control layer as well select pose from that that will give us a little button over here add character pose to selected layer there we go and then we'll get the character over there so we can select all the bits and pieces he's over there at the moment but now he's over there cuz we've just moved him and then we can move him again or we could make it a little bit bigger let's have that character bigger or we could duplicate it as well let's copy and paste and then we've got we've got two people in the picture or maybe one's a little bit smaller or at an angle or who knows uh you can even click on the little joints there as well as you see there remove that dot up there and then move the little elbow we'll zoom in move that one up there too and you know any little part of the skeleton you just move those legs out and have all sorts of funny things going on we can even scale it strangely too okay let's clean all this up and start over again with a little bit scribbling we'll get rid of that that's all nice and clean because another fun thing that you can do is just do sort of free form scribbling without any prompts now I understand that most of you won't quite be up to my skill level here so if you're trying to emulate this at home you may have some difficulty but don't worry just try your best and the program will try its best too as you can see it's sort of interpreting it over there and you know it's has not done a bad job there we let's give him some feet have some little paws up there there you go we've got like a kitten and it hasn't even got a prompt so you can just let it interpret your drawing and well see how good you are and see how good it is if you're happy with your generation you can click the copy the current result button and that will send your image over allowing you to carry on changing stuff such as I don't know coloring the eyes maybe you want red eyes or something let's make that big and uh oh not not quite that big there we go let's have some red in the eyes there we want it a bit more bit more evil now it hasn't changed very much but don't worry if you change the prompt strength it will eventually pick it up there you go so now we can copy that over it's got red eyes and I can draw within the lines now excellent brilliant got some slightly Rosy Cheeks there too as you can probably tell this is absolute bucket loads of fun and there's loads of other stuff stuff you can do with it as well this real time drawing just being one very small part for example it could be that you've already got an image maybe a face or a character or something and you'd like to somehow integrate that into your stable defusion art in that case this next video will show you a really quick and easy way to do that

Channel: Nerdy Rodent

Views: 29,747

Keywords: krita ai diffusion, krita comfyui, comfyui, generative art, ai art, making ai art, krita ai diffusion tutorial, krita ai diffusion howto, live diffusion, lcm, lcm lora, live ai art generation, Krita live painting, Krita live ai, Krita real-time, Krita ai tutorial, Krita ai guide

Id: AU8NDSBIS1U

Length: 14min 31sec (871 seconds)

Published: Fri Nov 24 2023