QR Monster Illusion Diffusion – Turn anything into AI art – Six QR Code Controlnet Models

[Music] foreign [Music] hey guys in this video I want to give you an overview of what's currently possible with the QR control net models you've seen some of it in the intro already and besides QR codes and text you can also do brand logos barcodes that from my experience are much harder to create and to scan than QR codes but also faces and illusions so basically any shape and form that you can think of if you want to learn the basics of control net and AI QR codes then I suggest you watch this video just keep in mind that you can also make those QR codes with a tile model in control net which you can see in this video the same goes for AI text or text Art you can either use the QR monster method which I cover in this video or you can use the line art method from this video the line art model also lets you do all kinds of shapes and in this video I used it to create a promo with a brand logo first I want to brief you on the different QR control net models as there's probably more models now than most of you are aware of and show you how to easily prepare input images to be used with the various QR models and finally we will have a closer look at the latest trend of the QR models that is to create optical illusions so right now there's a total of six QR code models for control net we have the two QR monster models version 1 and version 2 then we have a V1 and a V2 of the control net QR pattern models and we have two models called QR code conditioned what's interesting about those models is that one of them is made for stable diffusion 2.1 if you want to download the models you will find all links in the video description just make sure that you put them in the right folder and that if you want to try out all models always download both versions here are all models in the control net drop down menu and if you wonder how the model is different then have a look at this comparison where the prompt and settings stay the same for all images with only the model being replaced it's hard to tell which one is the best model that produces the nicest images but as you can see they all don't look too bad which is why I suggest you give all of them a try to prepare the images I am using a free tool called Irfan view where I first convert the image to grayscale and then go into the color Corrections where you can adjust the brightness and contrast until you are happy with how the image looks as you can see in this comparison the input image doesn't have to be black and white color works fine too but with the method you just saw in our example you can determine how much of the face is going to be visible on the final image using the models with any input shape is quite easy if you know the right settings for the sampler it is recommended to use a DPM sampler and you should enable the hi-res fix by the way in the latest versions of automatic 1111 you enable it by expanding the hi-res section by clicking on this Arrow so that it looks like you can see here then you can drag and drop your input image into the tool enable control net and check Pixel Perfect as preprocessor you can either select invert or tile resample and then you can select one of the six QR control net models these are the most important settings that you will be tweaking the most the control weight determines how strong the shape should be in the generated image while increasing the starting control step and decreasing the ending control step to more creative images and to a stronger integration of the input shape into the final image in this comparison you can see the difference of various control weight strings while every image has the same starting and ending control step values and here you can see difference starting and ending control step values with a fixed control weight knowing all this should allow you to generate your own images with a QR code control Nets no matter what you want to use as input image nonetheless we'll now have a special look at how to generate images with optical illusions as input if you want to use an easier tool than automatic 1111 or comfy UI there is now illusion diffusion which has reduced functionality but is therefore perfect for AI beginners you can either run it online from a website or install it locally on your PC the link for the website is in the video description and the local version can be easily installed via Pinocchio if you need help installing Pinocchio check out this video in the discovery tab of the tool all you need to do is locate illusion diffusion click on install confirm the name and click on install once again you can tell that compared to 1111 there are less settings and here you can set the strength that the input image should have in the final image you can upload your own Illusions or shapes as input or which is quite handy you can select a sample Illusion by clicking on it and if you then enter a prompt you can easily generate images if you're not happy with the output you only have this one slider which you need to increase if you cannot see the shape in the final image or if it's too strong you need to decrease the illusion strength and down here in the advanced options you can set how strong your prompt should be when generating the image and just a quick advice on how to get more input illusion shapes you actually have three options the first one being that you can just search and download shapes online or especially if you have a specific shape in mind you can create it yourself manually by either using online tools such as canva or any tool for presentations such as PowerPoint and lastly you can generate your own optical illusions with stable effusion for that you will have to download this tiny hyper Network and put it in the hyper networks folder under models then in the web UI you can go to the hyper networks Tab and click on the one we just downloaded as you can see it then gets added to the prompt where you can enter your own prompt and then click on generate you can then save those images and drag and drop them into control net where you can use them the same way that you are already familiar with you can also turn them into black and white as I showed you earlier so you can now do many different things with the QR control net models but what is still missing is a way to merge any of the shapes with existing images and there's already a few control Nets available for sdxl but none of the QR code ones are out yet and there are already ways to generate video but more on that soon by the way the videos I animated for the intro were made with cap cut I want to thank anyone who is contributing to the AI scene in any way and I want to thank all of you for watching this right now for your continued support if you make anything cool with aqr control net I would love to see it so if you don't mind please post the link in the comments in case you didn't notice not only the cards in this video like this end card but also all music and even my own voice is AI generated so check out my channel if you want to find out more I really hope you enjoyed this video and if you learned something new please like share or subscribe and I'll see you next time [Music]

Channel: Ai Voice Tutor

Views: 7,965

Keywords: Automatic111, Automatic11111, A111, A11111, Automatik 1111, Automatik11111, Automatik111, Logoart, Textart, Stable-Difusion, Stable Difusion, Stable Fusion, Staple Diffusion, Sabtle Diffusion, Staible Diffusion, Stable Comfy, Stable diffusion, Kontrolnet, QRCOT, AI QR Code, Barrcode, Barcotes, Illusion Diffusion, Illusions Diffusion, CRMonster, QRMonster, QRCode Monster, QRKode

Id: jvuLbw9SvjM

Length: 8min 36sec (516 seconds)

Published: Wed Sep 27 2023