Install Stable Diffusion Locally (Quick Setup Guide)

In this video, I'm going to show you how to set up and run Stable Diffusion locally if you're on a PC, as well as how to run it in the cloud if maybe you're on an older computer or a Mac. And once you've got Stable Diffusion set up, it opens up so many cool possibilities of things that you can do with AI art. And I can't wait to dig in. So let's go. Why would you want to actually have Stable Diffusion installed on your own computer? Well, the answer to that is there's zero limits with it. You can literally create anything you want with Stable Diffusion when it's installed locally, which is when you're using Mid Journey. They have some restrictions, some words you can't use. If you want to create not safe for work content, you can't do it. If you want to use words like blood in your images, they'll restrict you from doing it. Now, if you have Stable Diffusion installed locally, you can also use it with your own trained images. If you've watched my previous tutorial where I showed you how to actually train your face into the AI the CKPT file that it generated when you followed through on that process, you can actually use that model inside of Stable Diffusion and and generate as many images as you want with your own face on them. Having it locally installed has all sorts of other benefits. You can use all of the cool control. Net features, things like getting it to model your exact pose. You can add the Scribble control. Net where you draw a scribble drawing and it will try to draw a realistic version of that image. You can set it up with Image to Image, where you upload an image and it tries to create another image that looks similar to it. You can set it up with Instructpix to picks you can you upload an image and then give it text commands to tweak that image. You can set up Deforum on it, which is a video generator that uses Stable Diffusion to generate really awesome videos. If you have a local install of it, it's pretty much limitless to what you can do with it. Now, in order to install it locally, you're going to need a PC and you're going to need a pretty decent graphics card. However, if you have an older computer or if you're on a Mac, I'm going to show you an alternative that will get you all the same benefits of a local install. So the first thing you're going to want to do is go to huggingface CA and you're Going to create a free account. Just click up here on sign up. Create a free account. Once you have a Hugging Face account, you're also going to want a GitHub account. So go to GitHub.com come over here, sign up once again, totally free to have a GitHub account. So the next thing you need to do is you need to download a software for Windows called Get for Windows. So if you go to getforwindows.org you can download it. It's also completely free. Just come up here and click Download. Once this is done downloading, go ahead and click on it to launch the installer. Go ahead and click Install. Once it's done installing, go ahead and launch and click Finish. We're going to go ahead and minimize this for now. So the next thing you're going to want to do is create a new folder on your hard drive. Let's go ahead and call it AI. And let's just start with that. We're going to right click on this AI folder and we're going to select copy. Now we're going to open up this git bash that we created earlier. We're going to type cd, which stands for change directory. And then we're going to right click and we're going to paste in what we just created here. And then we're going to hit enter. Now we're going to come back over to GitHub here and we're going to search for automatic 1111. You can see that it shows automatic 1111 stable diffusion web UI. We're going to go jump to that there. And then over on this green button, we're going to click here on this arrow next to code. And then you can see a URL here. Let's go ahead and copy this URL. We're going to open up our get bash here again and we're going to type this command get, and then we're going to type space. And then we're going to type clone. And then we're going to right click right next to this and we're going to paste in the URL that we just copied and we're going to hit enter. Now, it's going to take a minute or so and it's going to start the process of installing automatic 1111, which is the user interface that you use to interact with Stable Diffusion. All right, so once that's done installing, we need to go download the Stable Diffusion files from Hugging Face. So we'll search out Stable Diffusion. There's multiple models. You can choose from 1.4, 1.5, 2.1 and 1.4 original. I still use the 1.4 original. Most of the images that you see where I've trained my face into it, it is using 1.4. So go ahead and select stable diffusion 1.4 original, and then under download weights, go and download the SDV1 Dash 4 CKPT. Now, this is a large file, so it will probably take some time to download, but go ahead and get that process started. You can see that it's 4 gigabytes, so just let that download. After you've got the 1.4 model, feel free to download the 1.5 model and the 2.1 model that you can add those into Stable Diffusion as well and experiment with the differences. Just keep in mind that each model does run about 4 gigabytes, and the more models you have, the more hard drive space you're going to end up taking up. Now, while we're waiting for that to download, you're going to want to install Python on your computer. So head ON over to Python.org come up here to Downloads. We'll come down to Windows and let's download the latest Stable Release, which is 3.10.10 as of the recording of this video. So we'll scroll down and we'll grab the Windows installer down here. Once it's downloaded, we'll go ahead and install it. All right, now that we've got Python installed, we'll go ahead and just close out of that, we'll open back up our folders. Let's double click into this AI folder that we created earlier. Double click into Stable Diffusion Web ui, look for this folder called Module Models here and open that Models folder. Double click into Stable Diffusion. And then under here it says put Stable Diffusion Checkpoints here. So remember that stable diffusion 1.4 file that took forever to download? Well, we're going to put it right inside of this folder here. Stable Diffusion Web UI Models Stable Diffusion. So grab it from wherever you downloaded it to and paste that into this folder here. Now, if you watched my previous video on how to inject yourself into the AI, if you selected this option here to save to Google Drive back during that training, you should have your own trained model inside of Google Drive. Now, you could toss that file in the same exact folder models Stable Diffusion, and that will also show up as one of your available models that you can generate images from inside of Stable Diffusion. All right, so the next thing we're going to do is we're going to hit back to Our Stable Diffusion Web UI folder again and we're going to scroll down and we're going to look for this Web UI Usure bat. If you double click on that, you're going to see it start to run a whole bunch of scripts. And these scripts are going to take a little while to run. So go ahead and just step away, grab a snack or something and let it do its thing. As you can see on my screen right now, it's showing ETA somewhere around 30 minutes. Once it's all done processing, you'll see down at the bottom here, it'll say running on local URL and then it'll have a web address here. Just double click on that, go ahead and hit Control C to copy that, jump over to your web browser, paste that URL in up at the top, and once you do that, you are inside of Stable diffusion. Running stable diffus 1.4, you can now generate any image you imagine. Now, there are a lot of settings here, a lot to play around with. Basically, if we come down here, we set our batch size to four, it'll generate four images. Every time we run something, I can go ahead and say a photorealistic image of Brad Pitt covered in tattoos eating a baseball bat and click generate. You see that it will run and start to generate the images. And this is what it thinks a photo of Brad Pitt with tattoos eating a baseball bat looks like. Now, when you first get in here, I recommend playing around with some of the sampling steps. Try different numbers, try some of the sampling methods. I personally really like some of the results I get with this DDIM as well. But try some of these different ones in here because you will get different results from them. Try different CFG scales. So for example, if I crank this up to 14 here, leave this one set on DDIM this time and generate again, you'll notice that the style of the images should be completely different. There you go. Completely different style of images. Now, I did mention that if you don't have a PC or you don't have a strong enough GPU to run Stable diffusion locally, I do have an alternative that's really, really awesome. It's called run diffusion and you can find it over@rundiffusion.com. now, this isn't a free option, but it only costs 50 cents per hour and you get to run 15 minutes for free the first time you mess with it. And the cool thing about run diffusion is it skips everything. I just showed you where you're installing it locally on your computer and it actually runs it in the cloud. You're using Run Diffusion's GPUs and their computers to actually run this automatic 1111 stable diffusion. Let's just jump into Run Diffusion real quick. I'm going to go ahead and click on get started. Now, I do already have an account with them, but if you don't have an account, go ahead and sign up for one. So I'm going to log in and you can see that they have three different types of GPUs that you could run them on. One's 50 cents an hour, one's 99 cents an hour, one's $75 an hour. And really the main difference between these three is how fast it's going to generate the images for you. You can see the $0.50 an hour one generates images in 3 seconds, $0.99 an hour, 2 seconds, and this one is actually 250 an hour and it will generate in 1.6 seconds. So let's go ahead and select the $0.50 an hour option. We'll jump in here. I have $27.70 in credits already loaded into my account. So let's go ahead and start a session. It asks how long should we keep the session running so that you don' leave it running overnight by accident ended up charging all of your credits here. I'm going to go ahead and leave it on for one hour. If you use less than an hour and you jump out, it will give you back whatever you didn't use. So it's not going to spend a whole hour's worth of credits if you set it for an hour and you only jump in and use it for 20 minutes. So we're going to go ahead and scroll down and click on Launch. It's going to start the little setup process when you get into it. It does take two to four minutes for it to basically turn on the computer for you. Once it's ready, it's going to play a little chime for us and we'll know it's it's good to go. So we can jump over to a different tab, watch a YouTube video, do whatever we want while this is booting up, and it's going to give us a little chime when it's ready. All right, so we got our little sound notification to let us know that we are ready to roll. Let's go ahead and hide that for now. And as you can see, this looks pretty much identical to the Stable Diffusion that we just installed locally. But what's really cool about this is remember, we installed stable diffusion 1.4 in that last one. And I said there's other models that you can play with 1.5, 2.1, etc. They have multiple of those other models that you can play with already pre set up in here for you to use. Let's go ahead and select this one called anything V3 safe tensors. Select our sampling method as DDIM here because I really like that. Let's set our batch size to 4 and let's generate a water color illustration of a wolf in the forest eating tacos. Realistic high definition trending on our workstation. Go ahead and click Generate and it will start generating my images. And it's actually pretty much just as fast as when I did it on my computer. And here's my watercolor drawings of a wolf eating tacos in the forest. This one has a human hand. Surprisingly, it actually has the right amount of fingers too. So that's how you set up stable diffusion in automatic 1111. You can literally generate anything you can imagine. There's no restrictions since it's either locally installed or paying for a cloud gpu. There's no limit to what types of images you can or cannot generate with this, and there's all sorts of different settings that you can play with inside of automatic 1111. The main reason I wanted to make this video for you is because I felt a little bit restrained on my YouTube channel because there's a lot of cool stuff I want to show you guys around. Things like Deforum, which is a tool to make cool animated videos with AI, and these various Control. Net plugins where you can get the exact pose that you want, and the Control. Net Scribble where you can scrib a drawing and it will try to model it. A lot of this stuff I haven't really been able to show you the full extent of what it's capable of because a lot of it requires that you have automatic 1111 either installed or running on a cloud server. But now that I have this tutorial, and now that you as a viewer know how to install this, I have a video that I can point people to to say, make sure you're running automatic 1111 in stable diffusion. And then here's some cool stuff you could do with it. It opens up a world of possibilities of what you could generate with images and videos using Stable Diffusion that you're just not going to get from Hugging Face spaces or Google Collabs. It's just so much more possibilities when you have it installed locally or running on a cloud gpu. This is going to be the first of probably many, many, many more videos where I dive into really cool stuff that you can do with Stable diffusion. So really exciting stuff in the pipeline and I'm excited to share even deeper what you can do with stable diffusion in automatic 1111 because it's just gonna blow your mind that you can do this and it doesn't take a PhD to figure out how to do it. I'm excited to share all that with you. If you like nerding out about all this stuff and you wanna learn about more cool AI tools, make sure you head on over to Future tools. This is where I curate all of the coolest tools that I come across. I add somewhere between 10 to 20 new tools every single day to this site and share them with you so you can check them out as well. If that's too many tools, click on Match Picks. It'll narrow it down to 132 tools that are just the ones that I play with the most. And if even that's too much for you, make sure you join the free newsletter. Because if you're on the free newsletter, I send you the TLDR of everything that's happened in AI that week and just my five favorite tools so you can find all of that@FutureTools IO. Thanks so much for tuning in. Hopefully you enjoyed this video. Hopefully you found it helpful. I'm excited that you're going to be learning Stable Diffusion now and doing some really really cool stuff with automatic 1111 with me. So I'll see you in some future videos. Thanks again. Maybe press the subscribe or like button. That always is helpful to me and I really appreciate it. All right, thanks so much. Bye Bye.