Install Stable Diffusion (A1111) in Google Colab | Stable Diffusion Basics

hello everyone this is me arit back with a new video and in this video we are going to talk about stable diffusion web UI so basically automatic 1111 so we'll see how you can install the latest version of automatic 1111 in Google collab and then I'll show you the basics of it like how you can convert a text to image and how by improving the prompts you can actually improve the generated image and then that generated image we are going to take to the image to image Tab and there I'll show you how you can further improve those images or modify them so all these things we are going to cover in this video now before we start the video If you haven't subscribed to the channel yet please subscribe to channel and hit the like button so that whenever we are going to make a new video on this topic you will get update and now without wasting any time let's get started okay guys so let's start so at first you have to come to this Google collab link so so I'll provide a link in the description you can just open that and you will come to here so you have to log in with your Google account and then you have to connect here so you can just connect here and also you have to go to view resources change runtime type and you have to select e4gp okay so it will take few seconds to connect so let's just wait for it so as you can see it's connected and the next thing is the first thing we have to do is we have to connect our Google drive with it and the reason is in stable we have to download a lot of models and some of them are at least like something like 5gb 10 GBS and something like that so it's better if you just save all the models in our Google Drive so that's why you have to connect to your drive so you just click here and it will be connected with your drive so it has been connected with our Google Drive next is we have to install our update automatic uh 1111 now basically uh in my case I already have installed so basically it will going to take update but but in your case if it is not installed it is automatically going to download all the things required and going to install uh automatically so you just click here and just wait for it so this part is done next we have to install the requirements again it will take some time so just click here and wait for it okay this done now we have to download the model now here I'm going to install the sdxl model but obviously you can install the models also in that case what you have to do you have to provide the links here or path of the model okay so if you have downloaded any model in your Google Drive already you can give the path here or if you want to download any other model rather than this models you can also give the link here okay but I'm just going to download the sdxl model and also if you a beginner or just starting you can download this sdxl model okay so you just select here and you just have to run this cell so done and now if you want to download any L you can also give the link here but that we are not going to do in this video finally if you want to install download any control net model you have to choose it here now I recommend you that you only download a cany one if you not sure what you are doing so this cany model is very good now you can also download all models but it will take a lot of space in your Google Drive okay so if you are beginner you are just starting download the cany one and you will be good to go at least okay for the at least for the very beginnner part i' would say Okay so so it's done and finally we have downloaded all the things we need now we finally can start stable diffusion okay so you just click it here and within few seconds we are going to uh get the link okay okay so as you can see that the stable division web UI has been loaded and here is the public uh URL so we can just click on it and within few seconds it will be loaded okay so as you can see here the Wei has been loaded and also the model you can see been loaded here so if you have multiple models you'll see those things here but here in our case we just only have that sdxl so we have it here next here we have to write the positive prompts so basically the features our generat image going to have we are going to have write negative prompts so basically the features our generat image should not have and then there are few other settings we are going to talk about okay so let's just take a small example so I'll just write here cute little toy owl something like that negative prompt and here I'll give it high quality and in the negative prompt I'll write low quality so I don't want my generat image to be a low quality also I'll write blur something like that now here in the generation sampling method I'm going to choose DPM Plus+ 2as you can also use Aller and LMS those things but I'm going to use this one then there is the sampling uh sampling steps now the more sampling steps you are going to have the better generat images is going going to be but also it is if you make it higher also uh the time required will be more okay next here we have the width and height so uh basically here I'm not going to change this 5 52 by 512 is fine for me okay next we have the CFG scale now CFG scale is the Lesser value will be the less importance it is the model is going to give to your prompt the higher the number will be uh the more the model is going to uh give importance to your pro okay that's the basic concept okay so 7 is a fine number but also you can play with 7 to2 7 to2 you can use and check okay and sampling St I'm going to increase a little bit so I'll just make it 50 okay and so that's it and now we can just click on generate so it's going to take few seconds for me to generate it so as you can see it has generated a cute owl image okay so obviously not that good okay but now obviously if you have to make your out output image is good your prompt should be very good so here you can go to different websites here we are going to get example of prompts for example here I'm going to copy this prompt and I'll just put this prompt here and the link of this website will be also in the description and I'll just copy this one I'll paste it here similarly you can check the parameters so steps 25 and sampler is Oiler a so I'll just choose maybe Oiler a and I'll just put it 50 no problem in that and CFG is 8 CED is this okay so CFG uh so CFG I'll just make it 8 okay but seed I'm not going to use the same seed because I want a different result so seed is basically if we put a minus one so it's a random seed that means every time you're going to generate an image it's going to be different than the last one but if you get give it a like fix seed then every time with the same parameters it is going to generate the same image okay that is the concept so here now I'm going to again generate it and let's see what we get so as you can see that this time we got a really nice owl okay owl toy I would say okay uh and as you can see it's not similar to this one okay and even if I if I just generate it again you will see you going to get a different one so let's just wait for a few seconds so as you can see this time also we got a new one and this one is also good so here as I think you have understood that the prompt has a very important role in generating this images so the better your prompt will be the better images you are going to get and also uh the generation this methods and sampling CFG all those things have effects but prompt is the most important thing okay now uh this is basically about uh text to image now I'll also show you another example but in this time I'm going to show you image to image so you'll take this image as a input and we are going to generate another image from this one so this icon you can just click and it is going to send your image to the image to image uh as you can see in the image to image panel it will just send it and now I'll just delete all this prompts and I'm going to give it a new prompt so the prompt I'll give it is that uh change the color to Red change the colored of owl to Red okay and I'll just give this one uh negative Pro I'll just give it low quality nothing else low quality and blur again and now again now here also you are going to get a lot of like what I can say a lot of settings here but yeah uh this one I'll just use d uh I'll use just DPM plus plus 2m car uh width and height I want to same CFG scale I just make it eight okay and and den denoising strength what is denoising strength is basically the more denoising you will add the more it will change the image the base image and the Lesser you will make it the less change it will do that is the basic thing okay now I'm going to generate it so let's see what we get so as you can see here it has generated owl type of thing obviously not good than this one but as you can see it has added red elements here okay so now what you can do is so I think it has haded like a high amount of noise here so what I'll do to decrease the noise I'll just make this Den noising Factor less so something like 6.5 I'll give it okay and now I'll just going to generate it so let's see what we get so as you can see it has added some kind of reddish color I would say in the overall image okay and now this time if you say that a lot of noise is not there so this image is not very much different from this one but is there some reddish tone is here I can see okay so similarly you can play with the settings okay so like you can check with different D noising strength different CFG scale and also like basically there is also features using which you can change a very particular part of this image not the whole part so those things I'm not going to discuss everything in this video so this was a very basic part of how you can use uh stable diffusion in Google collapse automatic 1111 so more videos on this topic are going to come very soon so I going to also going to cover the in paint in paint sketch all those things in some other video so stay tuned for those things and if you have question any specific question for this video let me know in the comment section okay guys so thanks for watching this video I hope you have liked this video and in that case please hit the like button and subscribe to your channel now many more videos on stable diffusion automatic 1111 we are going to make in near future so you will get update on all those videos and that's it for this video and I'll see you in the very next video

Channel: SPARKLERS : We Are The Makers

Views: 1,379

Keywords: StableDiffusion, TextToImage, GoogleColab, DeepLearning, AI, MachineLearning, Tutorial, InstallStableDiffusion, ImageGeneration, NeuralNetworks, GenerativeModels, ArtificialIntelligence, ColabTutorial, Text2Image, ComputerVision, GAN, Python, Code, Tech, DataScience, ImageSynthesis, Programming, imgtoimg, automatic1111, a1111, aiartwork, illusiondiffusion, deepfake, sdxl, sd, colab

Id: Q37eGFvMDbY

Length: 11min 20sec (680 seconds)

Published: Fri Oct 06 2023