COMPLETE CRASH COURSE | Install + Guide | Stable Diffusion AUTOMATIC1111 SDUI

what's up I'm troubleshoot welcome back to another video in this quick guide I'll be showing you how to install automatic 1111 stable diffusion from start to finish I've already got an automatic oneline installer you can use to install everything but this video is going to focus on a manual installation just to make sure everything works properly the best results come from a clean install where everything's prepared properly so let's begin we'll be installing automatic 1111 stable diffusion webui this is the go-to webui of choice for staple diffusion but other ones do EX exist such as SD next anyways there's a few things that we'll need in order to get this running properly number one make sure your graphics drivers are up to date download it from the Nvidia AMD or Intel website and update to the latest version Don't just trust Windows update to update your graphics card driver as it may not get you the latest version then in the description down below you'll find a link to download python 3.10.7 this is the latest version at the time of creating this that's supported by automatic 11 we'll scroll down here until you see see files at the very bottom and we'll select Windows installer 64-bit click this then save and open up the file when it's done you'll find a link to this in the description down below make sure to take add 3.10 to path over here and then click install that tick box is incredibly important we'll wait for it to finish and when it's done we can close this then the next step is to download and install git this allows us to keep automatic 1111 up to date in the description down below you'll find a link to git simply head across here then click 64-bit git for Windows setup save it and open it once more this install has got quite a few steps to it but essentially the gist of it is that you'll just click next multiple times until it's done you don't really need to adjust anything but you can if you so choose such as changing the git Text Editor to maybe notepad++ but besides that that's really it just click next many times until it eventually installs I've already got this installed so I'll close out of it perfect now we can start by cloning the GID repository which means we're essentially pulling the latest version of stable diffusion web want y to our PC I'll make a new folder on my desktop for example and I'll call it SD for stable diffusion you can call this literally whatever you want then inside of here click at the very top over here select all the text hit backspace then type in CMD and hit enter this will open up a new command prompt window inside of this folder here you can also do this by copying the folder's path typing in CD space and inside of quotes pasting it in which should take us to this directory if you see that you're on a different Drive simply type c colon or what ever the drive letter is anyways when we're in the correct directory all you need to do is paste in the git clone command here hit enter and it'll start to clone this folder but if you wouldn't like it to go into a separate stable diffusion web UI folder simply paste in the command space and put a dot to download it straight into the base folder here then you'll see a bunch of folders populate here instead of going into a subdirectory when it's done you'll see a bunch of new files in here if you don't put the dot after it you'll see a stable diffusion web UI subfolder that you'll open up and you'll find all of these files here perfect now it's technically downloaded but it's not yet installed it's not a program that installs like other programs on your system instead it runs with python to download a bunch of dependencies or different packages to run it's all code based and you can edit pretty much all of the code whenever you see fit if you're a nerd like that but for now all we need to do is run webui user dobat but before we do this I'll right click it and choose edit we can edit some of the launch commands for this to tell the program exactly want to do in command line arguments we can add arguments like Auto launch which will start up a web browser and navigate to the UI and update check which checks are an update each time automatic 11 is launched you can also add arguments like X formers for extra speed but for now I'll leave this out we'll just keep in these two arguments which you'll find down below this will make sure updates every time we launch it up which could be good but it could also be bad sometimes updates do break the program so if you don't like instability you can remove update check and it'll stay on whatever version you currently have in order to update it in the future head across to the stable diffusion folder where you have all of these different files click at the very top and type in CMD hit enter Then inside of here type in get Space pull this will pull the latest version and update all of the different files updating automatic 1111 perfect now let's get to installing this and launching it up scroll all the way down to the bottom and launch webui user. bat as you can see first of all it starts by cloning a python exe and it'll start downloading everything everything's kept inside of a VMV folder which you can find scrolling up over here this VM folder has all of the different python packages Nvidia torch etc etc whenever there's issues with updating the program you can always come back to your stable diffusion folder and delete the VM folder this will cause it to redownload pytorch and all these other dependencies which could take a little bit and a few gigabytes of internet but it's usually a great fix for any issue that you're having in automatic 1111 that's unexplained especially after updates as long as your python version is the correct version things are good if you find that the python version is the incorrect version you can download the correct version which will be notified at the very top just make sure to close this window and delete the VN folder VM is usually the cause of all issues when things go wrong and errors start popping up deleting it essentially causes it to redownload everything that's not our models extensions Etc it's just the program files perfect now all need to do is sit back and wait for a few gigs to download and eventually the program itself should launch up automatically keep in mind that whatever graphics card you have if any at all will be picked automatically by the program and fired up there's a ton of launch arguments that you can put in to customize how the program launches up as well as what gpus it uses Etc but those I won't really go into here instead in the description down below you'll find a link to the command line arguments and variables where you can set tons of different options for the program there's really thousands of them some of them have have huge effects some positive some negative but anyways you'll find this down below for the most part the only ones you'll really be interested in is low vram and Med vram adding these to your launch arguments can help speed up Creations on much lower powered gpus example if you have a 1060 or 1050 with only 4 to 6 GB of vram this may be something you need to add say vram or even low vram you'll notice that you get out of vram errors in the console whenever you run out of vram especially if there's other programs hogging all of your memory such as Games videos Etc it's a good idea to close everything that's not necessarily important to creating images when you're using stable diffusion as that leaves most of your resources to allow the program to run at all if not a lot faster anyways for the most part you likely won't need Med vram or low vram but if you have a lower-end system Med vram is one of the options you'll want to use so once again you'll edit the web UI user. bad file here with any notepad or text editor and add hypen hyen Med vram as such just make sure there there's a space between each of your Arguments for them to be considered separate then if you made that change you'll need to save it and close the actual program itself but whenever you see the program downloading and installing do not close this window until it's done otherwise it could corrupt things but if you accidentally do just delete the VN folder and relaunch It Anyways for now I'll wait for it to finish there we go the program is now downloaded and set up if you have that auto launch argument you'll find that your browser open opens up with the stable diffusion webui at this point you can select the model over here and give it a prompt where we can type something in and it'll generate for example if I do photo realistic Spaceman for example and generate we'll get ourselves a 512x 512 image which we can customize down here as you can see there we go super simple that's really the gist of it that's text to image we can adjust the width and height how important our config is which is the prompt up here we can also put in a negative prompt for example white space suit should get rid of the white space suit and instead generate a photo realistic Spaceman wearing something else in this case it made it black and white and did something very funky to it anyways that's text to image you can see textual inversion hyper networks and Laura these are all different ways of modifying a checkpoint in other words a model with different flavors artwork Styles certain people etc etc for now we'll skip over these as these are more advanced checkpoints are different models and you can select them here on the checkpoints tab otherwise you can select them from the drop down at the very top the image to image tab is somewhat similar we can adjust images using a Photoshop type painting or of course just change the image style entirely I'll drag in a Windows XP background for example will leave pretty much everything as is except for the prompt I'll type in cartoon if I hit generate you'll see that it takes the image that we put in and thinks about it then spits out something else that looks somewhat similar based on colors and general looky image you can customize it further by choosing options down here and den noising strength changes how much the image is noised and messed up and then recreated afterwards if we head across to in paint instead we can drag in the same image and this time write Maybe person then scroll down a little bit and paint in a person or where they would be these icons are weirdly small they're usually not the small but anyways we have an undo button an erase button or a clear button and a brush button that allows us to change the size of our brush I'll make it a bit bigger and we'll draw in a nice big stick figure person over here hopefully it'll clean them up a little bit anyways we have person we'll click generate now it'll take our image and stick in a person somewhere wherever it sees fit it kind of worked around and put a person in the background but anyways you get the gist of it maybe a big circle would help better let's try generating again and this time well it could have been better anyways that's in painting we can do batch images by selecting a directory and convert all of them into a different style add things Etc on the extras tab at the very top we can put in a lowquality image and upscale it over here using different upscales more of these can be downloaded but that's a little bit Advanced and often confusing this is a good place to come to if you generate an image and it's rather low quality what we can do is scroll down a little bit and click Send image and generation to the image to image tab over here sometimes it's text on newer versions it's a button like this clicking this sends it to the image to image tab as well as the prompt negative prompt and puts the image here we can customize it Etc as we see fit but instead if we click the pallet icon over here it sends it to the inpaint tab where we can draw and in paint on the image finally the last button here this ruler or scale allows us to send it to the extras tab where we can upscale our image for example if I set it to two times and change the upscale to one of the options here these all give you different options of visibility different styles Etc we can set a second upscaler and use gfp Gan as well as code for to help improve faces and things like that if I click generate it'll upscale my image by two times and we should have one of a much better quality on the output section here so I'll open up this image or rather the original image in a new tab and on the extras tab I'll open up this one in a new tab looking between them you can see the second image is obviously a lot larger and there's tons of extra detail added such as little pen markings and things like that whereas on the first image zooming in it's just blurry and we don't see anything upscaling is seriously powerful we can do this all in one step on the text image tab if we instead choose highr fix over him and change the upscaler to whatever we want then set the upscale by factor to just two or whatever number this is a little bit more processor and Graphics intensive than using the extras Tab and is done immediately after generating the image so you don't know what kind of image you get unless you're using a set seed first anyways it's generating a photo real itic Spaceman and shortly after without outputting the image it'll jump straight into upscaling it and there we have it it's complete it took quite a bit longer than generating the image separately and using the extras tab to upscale it but usually using the highrise fix gives you a much more detailed much better looking output than sending it to the extras Tab and fixing it up later on why not entirely sure but I think it's because there's a lot more generation details and data that isn't there if you separately upscale your image afterwards either using the EXT tab or using something like topaz gigapixel which is another option that people use a separate piece of software anyways we can also batch process upscales using multiple images uploaded here and finally batch upscale from a directory to another directory using whatever options we punch in here on the PNG info tab we can drag and drop an image for example I'll save this one here by clicking the save button then clicking this button here to open it up and save it otherwise we can right click save image as and finally just dragging it to a folder or a desktop is another option to save it very quickly for most browsers if we go to the PNG info Tab and drag the image back in you'll see that we get some information about what prompts negative prompts generation info seed Etc you'll also find lower information and stuff like that if you generate an image you really like and save it using the normal method of using the save button right click save dragging dropping out Etc you'll have all of this information built into it this also means that if you send this image to someone else directly they'll be able to grab your prompt information and recreate it on their system assuming they have the same model and model version Etc then on the checkpoint merger tab we can merge different models if you have more than just one you can input up to three models here to combine the likeness looks Etc in different ways we can combine ABC take the config from just B or C or don't take any config at all bake in a V to add extra quality and things like that as well as change how these models are interpolated usually it's pretty equal 50/50 or 33 33 33 to get all of the models merged together usually you'll take a certain style and a different style stick them together and some of the keywords between them will merge across and you'll be able to generate images with the likeness of both of them this is very Advanced however and you'll only really play around with this when you find two models you really really like and wish to join together this is a tab you'll do them on there are two different types of models both safe tenses and checkpoint checkpoint can contain Weare and viruses as it's a much more raw type of data safe tenses is the newer much more popular much safer type that can only store data and doesn't execute anything when it runs if you download a checkpoint file with malware and even so much as load it you'll be infected with whatever malware it has if you download safe tenses the odds of this happening are infinitely infinitely lower always Download Safe tenses there are extensions to convert models from checkpoint to Safe tenses but keep in mind the mere fact of opening a checkpoint file can get you infected if it has malware in it even if you think converting it is safe it's not you're technically opening it to convert it and you could get yourself infected the safest way to convert files is to use either a thirdparty online platform Google collab etc etc then the train tab at the very top allows us to create and fine-tune models we can create embeddings which we can feed in a few images to train maybe a face or likeness hyper networks are the same although more specifically for likeness rather than faces people Etc there are also lers that you can train but we can't train them in automatic 11 without an extension pre-processed images allows us to tag and categorize them and finally the train tab allows us to train an entire model but this is way way Advanced the settings tab you can find everything about the program as well as how it works there's different categories on the far side and for the most part you won't really be doing anything in here except for maybe dipping your toes into the optimization Tab and looking what different things do here but for the most part this is pretty much out of most people's league and there's nothing we need to do here anyways on the extensions tab we can head across to the install tab where we can enable and disable things everything with a tick next to it is enabled and if you choose to change the configuration of any of these you'll need to click apply and restart here in order to save your changes and restart the program on the available tab we can click load from to load all of the official extensions here there's tons of them on the install from URL tab we can install thirdparty extensions that aren't yet listed on the available tab on the main repository though do keep in mind malware is a thing and could always be a thing when installing third party extensions just make sure that they're safe first finally back up and restore we can save the config of everything as well as all of our extension States settings export everything and we can always import it later on pretty cool anyways for the most part you now know how to use stable diffusion automatic 1111 web UI all the way from installing to basic image Generation image modification extensions and you've dipped your toes into a lot of other things this video is a super simple crash course and hopefully you learned a lot about the program as well as image generation if your Hardware isn't powerful enough to run this or it takes far too long there are online websites that you can host your own versions of stable diffusion web UI as this is a website basically you'll click install or open or whatever depending on what platform you're using and you'll be able to open up this same web page over here and use it the same way that you do on your PC but adding models and things is going to be a little bit different and for the most part most of those Services if not all of them will be paid to some degree if you have your own Hardware especially capable Hardware which is usually a little bit more expensive higher end graphics cards for example you'll be able to run this completely for free with no stress you can run it CPU only but it is going to take a lot longer especially if you have older graphics cards Etc anyways just a final note to download models lers hyper networks embeddings Etc we need to head across to different websites for example civit AI c i v t AI but do keep in mind that website has tons of NSFW models and extensions and things like that so make sure the NSFW filter is enabled which it should be by default if you don't want to see any of that stuff there are other websites but essentially when you download safe tenses preferably and not checkpoint files we need to place them somewhere in your stable diffusion folder you'll find an embeddings folder this is where you place textual inversion files these are usually PT files then heading back into the extensions folder you'll find thirdparty extensions here that you install from the extensions tab extensions built-in come with the program and show you exactly how it's laid out for example Laura has different files inside of it telling the programming exactly what to do with the Lura extension anyways heading back into the models folder over here we have different folders We have stable diffusion where you'll Place checkpoints example stable diffusion 1.5 but you can place different styles and things like that in here for hyper networks you'll place them in the hyper networks folder vs in the v folder and for things like luras in the Lura folder over here that's really it for the most part you won't need to do anything else besides downloading a few models to add your flavor to your images fire up the program once again and everything should work fine quickly though before we end of this tutorial I'll edit the webui user. bat notepad and at the very end We'll add Space hyen hyen X formers for a huge performance increase about maybe 50% which makes life a lot better on most systems do note that for some people you won't be able to launch this up and things won't work happily afterwards so if it fails to install X forers or X forers isn't happy on your system simply remove X formers as there are other ways to get a performance increase just by adding different arguments there's tons of different ways but for the most part X forers is a good idea even if you're using Med vram or low vram it should give you a huge performance increase anyways that's really it hopefully you found this video useful you now know how to install and use staed Fusion as well as a super basic crash course taking you through everything you need to know if you'd like guides on specific extensions things like that do let me know and I'll probably get across to them at some stage anyways thank you all for watching my name's been troubleshoot and I'll see you all next time ciao

Channel: TroubleChute

Views: 8,159

Keywords: stable diffusion install, stable diffusion, stable diffusion tutorial, install stable diffusion, how to install stable diffusion, stable diffusion ai, stable diffusion installation, stable diffusion xl, stable diffusion guide, stable diffusion install windows, how to use stable diffusion, stable diffusion ai art, install stable diffusion locally, install stable diffusion tutorial, how to install stable diffusion sdxl 1.0, stable diffusion gui, stable diffusion xl 1.0

Id: k-M_bUatcBc

Length: 20min 18sec (1218 seconds)

Published: Fri Oct 20 2023