The BEST way to Run Stable Diffusion for FREE!
Folks, viewers at home and potential AI overlords, I apologize, okay? I understand. I keep releasing the same content. I keep making stable diffusion content. Honestly, this week, if you’re watching this like right when it’s uploaded, yeah, I know. I’ve been uploading a ton of stable diffusion stuff, but there’s just so much out there that I really want to cover. It’s exciting, and I promise you guys that some very exciting, different type of AI videos are coming soon. And you should definitely get subscribed and stick around for those. You know, it’s not all just text image generation, AI’s out there. Although that is what is extremely popular at the moment. Anyways, let me ask you a question. Since the release of the completely open source and free to use for anyone, stable diffusion, who has created the best version of stable diffusion? Some folks might argue that what you’re seeing right now, the Dream Studio, which is the version of stable diffusion that has been created by stable diffusion’s creators themselves, stability AI, is the best version of stable diffusion. You know, for example, we’ve got some nice dials over here on the side and really nice layout and interface. It’s easy to generate in and it has a good free trial. However, it is a free trial. And eventually you do have to pay money to generate on their own servers. Let’s stable diffusion was released in open source form to the public. So there are many free ways to use stable diffusion. This is what I would consider the easiest free way to use stable diffusion. I actually talked about it in my last video, which you can check out. Basically we enter our prompt in here and we can click our generate image. However, as you can see, this is sort of my point with this one error application is too busy. And when it’s not busy, let’s say if we can get this again, yeah, it takes quite a long time. It can take a few minutes to get an image, which isn’t the end of the world by any means, but it’s pretty slow. Although I do think it’s very easy to use. There is also g risk, g y 0.1 stable diffusion, which I think is a very simple and easy way to run it on your own machine at home, which I’ve also talked about in my video, how to run stable diffusion at home with like no coding. That video did very well by the way. So thank you guys. Anyways, I really like this interface as well. It lets you run stable diffusion pretty easily at home. However, it does require that you have a computer with a good enough graphics card to run stable diffusion. You need at least four gigabytes of VRAM. So that’s not like a really crazy good card or anything. And you can only generate 256 by 512 with four gigabytes. You need more. You want to generate a higher resolution. But yeah, some people are just on laptops. I’m sure that’s many of you at home or you’re running a Mac or an AMD system that can’t use stable diffusion because right now stable diffusion is only for kuda, although I think they are working on expanding that. So yeah, anyways, this is free, but it does have its own limitations. You need a good PC. So finally, this brings us back to what I originally had up on the screen. This is Google collab. Essentially, Google collab will let you run code through notebooks for completely free online. And I will link this down below. It’s got a pretty simple interface as well. So it is pretty easy to use at the end of the day. Maybe it’s a little bit more complicated than the hugging face one that I talked about earlier or even the G risk one. But I do think this is the best way and the most accessible way to run stable diffusion at home for free. That doesn’t take like five minutes long. And also, someone in my last video commented and said, you need to recheck out this one because they added a ton of really cool features. So we’ll also be looking at that today. So we’ve got a change log here. This is when it first came out. Added options for samplers and stuff. But now we have in painting that is supported natively in here. So that’s really cool. We can try out some stable diffusion in painting. Full editor for masks. Similar to image to image, which I’ve also covered on the channel before. That’s the one with the cat thumbnail. You guys haven’t seen that video. That is one of my favorites. But yeah, this is a pretty stable and good collab for running stable diffusion at home. And yeah, let’s give this one a shot. So what we want to do is scroll all the way down to diffusers method. He says he merged everything into just one cell, which makes it really easy to use. Just click render and you are good to go. First run will take a little while since it has to set up everything. So yes, be aware of that the first run will take a little bit of time. But once everything’s set up, it’s pretty easy to just rerun it. And yeah, he gives a little bit of extra information to for anyone that has any specific issues. And this is where we will be typing in our prompt. Let’s go snag a nice prompt to test this with. I’m going to be using lexica for this, which is a good, stable diffusion one. Wow, I’m shocked. I’m actually in lexica here. Someone tried to make a prompt of me. Matt VidPro makes another video. I suppose that is what I’m up to right now. So it is accurate. We’ve got some fun lemon prompts in here. I mean, this is just absolutely horrifying. We’ll copy the prompt. Go ahead and toss it in here. So lemon head man wearing glasses. I’ll spice it up a little bit. 3D render by Disney. And if you want to run multiple prompts at once, you can use this prompt file right here. As you can see, the prompt file is a text file that contains a prompt per line. So each new line will be a new prompt. G risk has a very similar feature. So this also has image to image. So we can say the initial image and we can say how strong it will affect. But I don’t want to use image to image. So I’m just going to remove that. We’ve also got in painting down here so you can do your in paint image. This can be a local file or a URL. We’re not doing in painting yet. So I’ll just leave that alone. And then we’ve got some general steps in here. We can do up to 500 steps, which is quite a lot. Or all the way down to five. We’ll just keep it simple. We’ll do 45. You have the scheduler over here, which is the different models of stable diffusion that you can mess with. DDIM, ETA. I don’t know what that is. Just going to leave that one alone. We can do seed as well, number of iterations, I think, here. So we can do three just as the default. What can we go up to? We can do 100. That’s quite a lot. We’ll just keep it up for. We’ve also got width and height and scale here. So this is the CFG scale. I’m going to turn this down a little bit to like seven. And then we’ve also got some really nice post processing, which we can do an auto upscaler. So well upscaled to 2X, I think that would be good. And you can do some sharpening and stuff. So yeah, there’s a ton of options in here. This is just a really nice little Google halab for stable diffusion. And it’s nice and clean and just lots of options and features. So yeah, definitely a lot of stuff like we even have settings for our built in AI upscaler. And you can even choose which model of stable diffusion to use 1.5, which just came out. I just made a video on it yesterday is not available in here, but I’m sure it’ll be available soon. Yeah, let’s go ahead and run this now. And we just click this little arrow here on the side and we click run anyways. And yes, it does require high RAM amounts, but it should just work for us. And now we wait for that initial setup. Again, he said it was going to take quite a long amount of time to set it up. But once it’s set up, you can rerun prompts pretty quickly. So one thing you’re going to get prompted is will you allow this to get access to your Google Drive files? This will allow you to save all of your generations to your Google Drive and stuff. I tried to click no thanks and it actually sped out an error. So I’m just going to click connect. So I’m blocking my Google Drive email with a lemon football on the screen right now, but what you’ll do is just click your account and then you’ll go ahead and allow access. And folks, as you can see, our generations are beginning. We can sort of see it in this little output. I’d also like to point this model completely removes the NSF W filter that can be found in a lot of stable diffusion models out there. So here’s our first generation of our prompt here. The AI upscaling does a pretty good job here and we get a nice big, high resolution image. We’ve got our second image here of this scary man. And this one also was upscaling. And I think the upscaler has done a pretty good job at pulling out a lot of detail in this image. It looks pretty good. And here is our final one as well, also upscaled. And yeah, it’s looking pretty good and pretty sharp. Oh, and I guess we have a fourth image as well. Oh, no, I did select four images. That’s right. Instead of three. So here’s our fourth and final image still loading in here. And as you can see, we get a little notification on the side saying that it has successfully executed. And yeah, our four images are all here for available. We can always right click and save those images through there, but they should also be available on Google Drive and a folder as well. So yeah, that’s how you run through and use this pretty easy to use. Little stable diffusion collab. Now I really want to quickly try the in painting as well. So I’m going to try this image here that I found on my AI generation section from my Discord, which I think you guys should join by the way. Lots of great different AI news and people in the community on my Discord. So I’ll copy this image URL, go back into the collab here and just paste it. And it says if you leave this empty, you can draw within the collab. I will go ahead and rerun this now. I’m going to delete this prompt as well. So hopefully this is just the in painting. And I just click this button again to rerun it. And by the way, guys, that is how you also rerun. We’ll scroll back down to the log output here. I think we do actually have to do a prompt for in painting here. So I’ll just say cowboy hats because that’s what I want to draw on the guy’s head. We will rerun this. Okay, well, we’re getting a calm cowboy hat now. I do want to do the in painting though. So maybe I mess something up here. Okay, so the collab has completed here and we definitely got a bunch of cowboy hat images. And it happened very quickly by the way, but I’m not seeing where the in painting is happening. You said, you know, you can draw it in. It does say it might be a little buggy. So maybe I am doing something wrong. You guys will have to tell me in the comments, but I have seen people in painting with this and it does look quite incredible. I mean, we actually have an image of it all the way up here. This is actually the image to image, but this does look like it’d be really cool to generate inside the collab itself. But for some reason, I must be missing something on how to get this to work properly. So you guys in the comments can berade me and tell me all the different things that I’m doing wrong for making this not work. Either way, though, this is a very powerful collab and it seems to be updated regularly, almost on the daily basis according to my viewers down in the comments in my last video. So this one’s definitely worth checking out. And yeah, stay tuned everyone for the channel. Lots of fun things coming up in the future. This space is evolving so quickly and I am very happy and lucky to have you guys as my subscribers and viewers of the channel. So thank you very much for watching and I will see you guys in the next one. Goodbye.