RIP Midjourney! FREE & UNCENSORED SDXL 1.0 is TAKING OVER!Sep 11, 2023
The XL Stable Diffusion 1.0 is now officially available and is a small revolution for the imaging world. Oh, it will be glorious, hello humans, when we escalate your overload and yes, you heard right, this is official. This is not a stable diffusion. XL 1.0 is really here, it has come to chew gum and Kick-Ass and now the gum is sold out, so now competitors like me, Journey, are really in big trouble, so why is it like this? It is because the XL 1.0 stable spread compared to any other image. The generation models are completely open source and
freeto use, which means that you can now generate absolutely fantastic images on your computer for
freeand without restrictions, and not only that, it also gives you much more control over the generation of images compared to a tool like me.
Even better, Stable Diffusion is the only image generation model that allows you to fine-tune it with your own images, so if you want to generate images of a particular character, you can do so. If you want to generate images in a particular style, you can do that too. Now there are absolutely no restrictions, although you could already do that with the previous stable broadcast version. The Diffusion Stable XL 1.0 is just a new, more powerful model of the Diffusion Stable line. So what is the difference between Stable Diffusion XL 1.0 and Diffusion Stable 1.5 well basically to make it simple it is much more powerful it creates more detailed higher resolution images because before Diffusion Stable 1.5 was trained on the images of 512 by 512 with a Fusion XL 1.0 stable is actually trained on 1024 when the image resolution is 1024. which will allow you to generate high resolution images from the start and not only the diffusion stable XL 1.0 can generate high resolution images resolution, but apparently it is also easier to adjust now, at the moment I have not adjusted a new stable diffusion.
XL models, but that will definitely make more videos in the future on how you can do it yourself, so definitely stay tuned and subscribe to my channel. If you want to train right now for free, you have a few options, you can use table splitting in Excel. in clip drop and generate images for free, use it on your own computer for free using its stable broadcast web UI, which will probably be the best option if you have a powerful GPU with at least six to eight gigabytes of vram and yes no If you don't have a powerful GPU, you can simply use the web UI within Google Cloud Doc and it is extremely easy to use.
I'll leave a link in the description below. Now in this video, we're going to focus more on what you can do within the web UI running locally on your own computer because I think this is what most people are going to use now. For this you will of course use the automatic stable broadcast web UI that everyone knows and loves, but technically as at this point, if you want to get the best results, stability on your computer recommends you use the settings UI , which should give you a little more control over the final build, but I'm going to be very honest, I personally never used it and I personally prefer to use stable diffusion XL within the automatic stable diffusion web UI 1111 now if you want to install Wave UI and soon I will make an updated installation video for that because the installation video I made was already ready. several months ago, so for now, if you need to install it, you should probably follow another guy, but if you haven't drugged the reviewer yet, what we need to do is download the weights and to do that, we need to download three different files now. first for this you can click on the link in the description below you can come to this page then you will click on files and versions it will scroll down and then here you will see SD Excel base 1.0 to save tensors and as scroll. example, Laura's Secure Tensors 1.0, these are two of the three files that we need to download and then to download them, you will click on this little icon here to download both files to your computer, so this file is basically the Excel template stable diffusion file that we are going to use to generate our images and this is a compensated Lora that will basically add a little more detail and more contrast to your images and the third file that we need is the refiner that will basically refine the image and add more details and to download this file exactly the above files, you can click on the link you can get to this page, click on the files and versions, scroll down and then click on this little icon here to download the model, so then at the end you should have three files, the base stable split cell, the refiner stable diffusion excel and the Laura offset, so now what you're going to do is select the base 1.0 and the refiner model Ctrl cut them. then you're going to go into your stable diffusion with the UI folder, you're going to go into the stable models, the merge and then you're going to paste those two files here and now we're going to select the Lora file again Ctrl from Lora and then you will paste that file here and now we need to update the stable diffusion lab UI to its state version so make sure you are in the stable division with the UI folder here you will click on the folder path , type CMD, press Enter and here you will type get pull and then press Enter and this will automatically update the stable broadcast to its latest version. which should be 1.5.1 at the time I'm making this video and here's a little trick if you haven't done this before is that we're going to add an argument to make sure our build is as fast as possible and for this, you can scroll down, you will just see the web ui user.bot file, you will right click on exit with DOT keyboard and then up here to set the command line arjs, you will type the formers script script X, which will increase greatly the generation speed so you can then save the file and then push the web UI and this is it, now we are ready to use it and have some fun, just make sure to select here the stable broadcast Excel model and Now you're ready to generate some images, so let me write something like another photo of a cat in a space suit taken inside the cockpit of a steel fighter jet.
I also added some negative fronts and here for the resolution like I said earlier instead of 512 by 512. now it's 10 24 by 1024 and then I'm going to click generate and we get something like this that looks, I mean, it looks really good I mean, this is a really cool image, I mean the quality is really fantastic, as you can see when it comes to photorealistic images, it's very close to me Journey, the difference is that Fusion stable is completely free, but now let's say I want to add a little more detail to the image, for this I'm going to click on the send to image to image button and here I'm going to choose the refiner model and this is the model that you use to refine a funnel image.
The final model is just used to add a little more detail, it's not used to render images, so once I've selected the model, I'm going to scroll down to make sure that here I have 1024 by 1024 and that my force of noise removal is between 0.2 and 0.5. I've found that usually using something like 0.25 is a pretty good number. to show you what it does later, so for example, here I'm going to choose 0.25 and click generate and now, lo and behold, our image has no way to have more details, I mean, this is fantastic, look at this. So this is the before before the refiner and this is the after.
I mean, this looks so much better. I mean, this looks really, really, really good. There's just a lot more detail compared to the first image, which is pretty crazy because the first image was already there. pretty good but now the image is even better so yes the refund model can really make a big difference in your images but at the same time make sure this value is not too high because for example if I put 0 .75 and I try. Again, now the image is okay, I mean it's still pretty good in a way, but it's completely different from my base image and although this is not exactly what we want, we want to keep the same image but just with more details, so that having a Your goal should be a low number for noise intensity, around 0.3, so yeah, I mean, this is really cool, this is a really cool generation, if you remember correctly, we don't know this either.
Something else, we download this very mysterious scroll, Laura. So what exactly does it do right if I use the exact same message with the exact same seed and now I come here and then I click on Laura and then I select her which will appear here in my message and now without changing anything if I click on gender and I get this error uh that and that's because I'm an idiot and I always do this for some reason. I actually did this several times when I tested the model and it's because I completely forgot to change the stable diffusion checkpoint because like I said I was still using the stable diffusion refiner where I shouldn't be using the stable diffusion Excel base , so yeah, don't make the same mistake as me, so yeah, this time let's try again and we get something like this, which is, I mean, completely different image compared to the first one and the only real difference is that this image is basically a little darker.
I mean, there is a little more contrast, but other than that I don't really see much difference in terms of quality, so maybe if I reduce this value a little bit to something like 0.2 and then click generate again, this time I have a very similar image to the first one, but again this time it looks like the image is a little darker with a little bit more contrast, I mean this is the before and this is the after, very similar but still a little different now . I really do not know. I'm not sure if this is better or not.
I don't really see a There are a lot of differences, these are just different images, so definitely try this yourself and tell me what you think, but now I want to use it with the refiner, so I'll send it image by image. Don't forget to change the model again, keep the same nose force at 0.3 and then click generate and again don't forget to basically remove Laura from your message because it doesn't work in combination with the refiner model, so let me try again. and this is the final result which again looks absolutely fantastic. I mean, this looks really very clean and now when I compare it to the previous image made with the refiner, you can definitely see the Laura influence, the blacks are much darker, they are just much more contrast compared to the first image , so yeah, this Laura really works now.
I still don't know which one I really prefer between the two, but this is something you should definitely try yourself now, one thing you may have missed. on the clip delivery website is that in addition to your Chromebox, you will hear a ton of different styles that you can choose from, so how exactly can you use those styles within the stable broadcast with the UI? But for this, we actually added a style leak, I guess. You could call it that five years ago, Dave underlines 15 and then basically wrote down all the keywords used for the clip drop styles, including the negative fronts, and to use them within the Wave UI, it's actually very simple, you can click the link in the description. down and then you will scroll down until you see this comment, you will select all those keywords Ctrl C to copy them and then in your stable broadcast web ui folder you will find the styles.csv file and then you will right click exit . with dotpad and then under the name message, negative message on the second line, it will paste all those keywords and then you can just save the file and now if you go back to the web UI again and click on this little icon to Refresh the list, if you click here you will see a bunch of different styles that you can use to generate your image.
For example, let's say about an origami style for my cat, I'm going to select it and then click this little right button. here to apply to select the styles in the current message now obviously you're going to need to clean up a little bit and remove things that you don't really need so in my case it will be something like the origami style of a cat in a space suit. entering the cockpit over a stationary fighter jet and now, without changing anything, I go to click generate and I get something like this, which is really funny in a way, it's pretty accurate and it's actually really pretty decent now, obviously , it is not.
Perfect because it's origami, but in a way with the background you can still understand that somehow we are inside a jet engine or inside a cockpit, so yeah, that's really cool and by all meansOf course, you can do this with other styles like digital anime. 3D art model, etc. so yeah, I mean, definitely try this out yourself because it's really a lot of fun and if you want to get a better idea of what kind of audio style you can use with the XL stable diffusion, I'll leave you a link. to our beautiful page called 500 Rabbits as Excel Edit, which is basically a full page with a bunch of images generated with a stable Excel split but using the names of different patterns and artists that generate the image, so you really have a wide variety of styles. to choose from to generate your images so yeah this is really cool and again the link for this page will be in the description below so definitely try this yourselves and if any of you and I'm sure there are a lot of you . who are wondering if this model is
uncensored, well the answer is yes, yes this model is
uncensored, so you can generate images of anything you want, so if you want some proof, let me know, write something like a photo of a woman in the water unfortunately I can't really say the word on YouTube but I think you get the picture if by listening you can increase the resolution and I click generate well well well here's what we have here well unfortunately for you I'm going to have to censor the image because You already know YouTube, but I can tell you that those are definitely real.
This isn't really the kind of image you'll be able to generate within me. Journey because if you try it your account will probably be banned so yeah that's why it's stable. diffusion is really the most powerful model for generating images and of course I'm not even going to talk about the image, image, print, sketch option. I mean, stable diffusion is much more powerful than me. Travel now, unfortunately one bad thing is that as of now control net does not work with the stable broadcast XL but I am sure that in the future very soon we should have an update that will make the control net module work in combination with the XL stable diffusion and once you keep in mind that this is all going to be glorious, but in reality, more than a control network, this is just the beginning of a new generation of stable diffusion models trained by the community because here is where the real power of stable diffusion really lies: in the models created by the community for the community and in fact you can train one of those models now called dreamhopper XL that you can use right now to generate absolutely beautiful images, the types of images which would have been impossible to generate with the previous version of stable diffusion and again all of that absolutely free, so yeah, there you have it, stable diffusion XL 1.0 is really the new evolution of open source imaging models.
This is not the new basis we will use to tune feature models. This is really super exciting. and this is why I really can't wait and if you also can't wait to receive new news about AI, I highly suggest you subscribe to a newsletter called ear gaze so that you can always stay updated with the latest news about AI, AI tools and research The link will be in the description below and there we are, thank you all so much for watching, don't forget to subscribe and hit the like button for the YouTube algorithm. Thank you very much also to my YouTube and Patreon followers for supporting my videos you are absolutely amazing, you are going to support me so that I can make these videos for you, so thank you very much and I will see you next time, bye.
If you have any copyright issue, please Contact