Realtime AI Image Generator with SDXL Turbo. Stable Diffusion, ComfyUI, Automatic1111

let's try see how fast can Ender right now so we're going to do juristic and notice as I'm typing it's already rering futuristic spaceship while painting did you notice how fast is as I'm typing it's because we using the new um turbo sdxl that's what call actually this is ADV diffusion uh distillation it's a specific news how train model was developed um relatively it's actually not totally new but it's new with a clip installed in the game inside so it's creating excellent models on the Fly and what I was meaning on a fly we have this sdl XL turbo and it's creating 512 by 512 so it's creating actually quite big size and again if you notice right here it does literally on a fly okay we're going space bottle and we can do all of these anything for example let's go ahead um cute puppy there you go instantly you can notice how it work very well as train model so let's go ahead and see how you can install this and as example we will be running this on the this on a com UI as well in automatic elimin 11 on stable diffusion so we look how to install on both of those platform and it's surprising very very easy to do this if you're not familiar with comy UI please watch my other videos how to install and you can install directly confy UI or you can use other ones like Matrix UI uh Matrix AI or Matrix stable diffusion sorry I am don't even remember Matrix stable diffusion to install it so it's very very easy to do okay let's start it first if you are geek out about AI like I do I would recommend to check these papers that was published about what is using it's ADV veral diffusion installations so it's kind of like teacher student and you can go over with a very nice explanations how the algorithm work what they done how the model was trained in this case and uh what result can provide it as well how they all work machine I know it's going some and numbers it's why I say if you geek out on this you can look on this and also in this papers you'll notice it's have some limitations this models primarily it is limitation on training so what can do in realism because it's going so fast but it is incredible step forward in this technology so this is the first things and links will be provided down below for you in descriptions next what you want to do you want to download this model this is very important it's sdl EX L turbo 1.0 fp16 safe sensor the model and this is a 6.94 gab so about 7 GB downloaded and same model will work for conf UI and stable diffusion you want to take this model and place inside your um stable diffusion web models and you want to go to stable diffusion folder for the models and just put here I already have a little bit too much of the models inside of this and that's it next what you want to do is run your installation and automatic 1111 right now select the scl the properly model in the top drop- down box next in generation and this quite a bit important because it's how model was trained so for the best result you want to do this sampling step one and this is what make it so fast because usually you may have it 50s 70s sampling meod so so it's can go over it in and over to create it in this case just one sampling method so it's just only run once which is incredible how it's work so we have it one we want to have it width and height 512 x 512 it what was training on size UL a our simpling method and CFG scale we want to set to one and that's all next what you want to do just go ahead type prompt for example let's go um type bird cute b or whatever you want it generate and there you go you can see how fast it is we can also for example let's say batch we'll set two maybe let's go 10 10 10 or 12 12 images let's go ahead do this click generate and you can see how fast it's generating 12 images again it's a 512 x 512 so this is incredible of course if look here you can see it's a very fast generating them on our server and right here we have it all or our this images that we just generated I think it's incredible so it's very easy to do with this U next if you want to use the with confy UI it's um you need perform couple steps okay one we want to be sure our comy UI it's have properly model installed same TBO what I done in confi folder I just spits five for example went to config portable conf UI found my extra model path it is spey additional path instead downloading this model and you know copy and everything I went the example renamed to the yaml and inside all what you need to do is just go right here where it says base path spit spy specify path for the stable diffusion webbi it's where my models located so I just leave it everything just modify only this path and say it's on this case the um my comp UI will will share same models as I'm using with atomatic 1111 the next what we want to do it's downloading the um no manager right here and this is a custom node I'll show you a second how it downloaded but what it does it's allowed us to install missing custom node just in case if you're using and you don't have them for this going to your comi folder going to the custom nodes click on this notice I already have it my installed but you can do same things just on the top in your file manager go on top right here type CMD or command manager so it will open for us right here command manager next what we want to do it is go and type G clone h HP again the command will be down below for you so you don't need to worry about this type this and it's take the repository and put it inside that folder if everything successful inside a custom folder you should see now comy UI manager after this installation you do need reinst re rerun relaunch your comy UI of course if you don't have all this nodes you want to install and the easy way to do is use it with a reference inside the image or Json file I will provide link down below and it's special thanks to stable diffusion r.com for providing additional knowledge and information and Link will be to the Json file all what you need to do just grab drag and drop and you will have this not settings inside your UI you will see this appear the schematic if some of these nodes will be most likely like this one will be red for you it's mean node not found then you want to go inside the manager and inside the manager click on install missing custom nodes all what it does it's come up right here showing you if you're missing any nodes and you just go ahead click install and that should appear if they all successfully downloaded in your machine you'll should see them appear inside here um if needed again I just recommend I use this all the time those nodes and they also you'll see them appear if they missing inside your custom nodes subfolder around here now you can use your installation com UI and for this one just go ahead and type like cute Poppy and let's go ahead press qu prompt and you can see it's almost instantaneously render of course the speed of rendering will be depend on what video card you're using I'm currently running on RTX 3090 with uh I think 24 GB Ram so if you have a little bit slower card this maybe a bit slower but generals run very fast and one other things the confy UI can do that you cannot do with a automatic 1111 yet it is real time rendering and what I was meaning by this if we going on kind of size right here click on extra options we want to go down below enable after cor so this is will be monitoring for our prompt to launch we need Press Start q and you'll notice in our the terminal we can see right here it's constantly watching for the keyboard so let's go ahead and start type and I'm going to raas and says for example um border collie dancing in the rain and notice as I am typing how it's change very fast so it's kind of very cool and you can do for example um futuristic spaceship while painting you can see how it is creating on a fly very fast as you're typing in rendering so fast remember it is Will depend on your video card but General I find it's work extremely well okay this is how simply you can install in automatic 1111 in stable diffusion you can use it as DXL just be sure keep it the sampling steps the um sampling method and CFG to the properly settings you can play around with them see what is happening what the result will be and as well in comy UI allowed you a lot of flexibility and real time render almost as you're typing so you can preview this well if you like it be sure you subscribe like and share I'm going to actually experiment with this to create animations see how well it will perform so be sure stay here to watch those videos when we'll going Explore More with this new Fast model because it will change in a way you can process videos well thank you again have a great day and have fun creating bye

Channel: Vladimir Chopine [GeekatPlay]

Views: 3,471

Keywords: Geekatplay Studio, Vladimir Chopine, Digital art, How to, Learning Photoshop, 3d tutorials, Digital compositing, How to do in Photoshop, How to do compositing, How to do in 3d, AI art, MidJourney, Stable Diffusion, Dreambooth, Dall-e, Free resources, Free learning, Digital art for begginers, Free tutorials, artificial intelligence, Digital Photography, Photography

Id: Y-0mQIiRlBM

Length: 11min 29sec (689 seconds)

Published: Wed Dec 20 2023