Hello and welcome to this video in which I want to exchange some life time against knowledge again. I was already lying flat again, so I have to pay a little attention to my voice, but nevertheless I'll show you something cool again today and that is the brightness control net. I've been playing around a lot with it lately and you can already do very, very cool things with it. And before we start, you do the whole thing here. This is a control net that you have to download. I will of course pack a link in the description and if you are on the page here, you go to files and versions and then you have to download this Diffusion Pytorch model.savetensors. This is 1.45 GB big, it hurts a bit, the record is running fuller and fuller, but it doesn't help at all and you download it in your ComfyUI folder under models and there is a control net folder. I have made subfolders for me according to Stable Diffusion 1.5 and SDXL, but you just need to pack it in here or create such folders yourself and then the whole thing is where we have it up here. You have to rename the file when downloading because it of course comes under this name Diffusion Pytorch Model Savetensors. I just have taken the name of the repository for the file name up here and then pasted it in and downloaded it and then the whole thing had the right description. So that as a start and what can you do with it now? Brightness Control is intended to include a brightness or also darkness of a picture in the image generation and you can see that best in examples and that's why we'll start now. I'll say Load Default here, then we have a Stable Diffusion Default or ComfyUI Default setup. I always like to make the node a little bigger because you can't load the checkpoints otherwise. I already choose our Epic Realism here. What I always like to do at the beginning is to make a reroute from the VAE here and connect it to the back here because, simply because of that, we can say later that we want to add a Load VAE. We don't need that right now, although we do actually need it because Epic Realism doesn't really have any baked-in. We can then connect it there because, to be honest, we can change the model here, we don't need another node. Clip rather unlikely. If you work with other clip models or something, then yes, you are very welcome to put a reroute in here, but the VAE, you might want to be able to switch it during the image generation and if you have a lot of VAE Decodes and VAE Encodes at the end, then it can be very annoying if you suddenly want to reconnect everything from a VAE loader. That's why you pack a reroute in here at the beginning. Then you just need to switch here and all the connections will simply remain after that. So good. We take a few more settings and I would like to change the sampler and the scheduler here . Then I would say let's go to Fixed here. We'll adjust the prompts right away. I think I don't need a safety prompt here at the point, but what I would like to do is, I don't want a Latent Empty Image, but the Stable Diffusion 1.5 aspect ratio from the ComfyRolls. Because I want to generate a 4 to 3 landscape, I'll take 4 to 3, 3 to 2, we'll take 4 to 3, we'll put that up here as Latent, we'll push that a little bit to the right and I would like to also switch to an Image Save. That means otherwise I think Save Image, right? No, that's the right one. How was that? Swoops, because here I can easily indicate a folder and simply say Brightness. Then I have the pictures that we are going to create here right now. Directly in the right folder for thumbnails. You already know that. Here we just make a preview image. We'll make it a little bigger so that we can see something right away. And then we'll start. First of all, I would like to have a beautiful landscape with mountains, clouds, blue sky. Then I would like to have trees, green meadows and houses. And down here, just so that it gets a little nicer, let's say horror. The prompts have already beautified that very much. I'll see if we've set everything right here. That looks pretty good. We'll just give it a try and then we'll see if what we generate as a base image is already good. That doesn't look wrong. We can leave it that way. We'll work with that. Well, Brightness Control. We have to clamp the whole thing in between here and we'll start with a ControlNetApply. As we know from ComfyUI, we want to hang the Conditions in between or the ControlNet in the Conditions. I'm just thinking a little bit here that it stays a little clearer. So we need a ControlNetLoader, of course. Here we can already take our Brightness ControlNet. There it is. We can pull the note a little bigger so that you can read it well here. And of course we need an image. And to generate the image, i.e. an image with the brightness information, I take from the ComfyRoles under Graphics -> Text -> Draw Text. We can use this to draw text. I would like to show you how we can render text into the image, but so that it looks like it just belongs to the image. And then we have to make a few adjustments here. I'll take Impact as a font and so that we can convert that into the right size, we'll store that once. And now we can also take the width and height down from the aspect ratios up here and take that as an image that we send into our ControlNet so that we can see what we are doing at all. Let's do Preview Image once. Turn off the whole shoe up there for now. So that we can now create a nice text image here. And for that I go to AutoQueue here. Press Start. Yes, that was stupid. We need that, of course. That's a dependency. So now we have AutoQueue. First of all, let's say we want to have the font color white. Then we see our text. And now I'll write in here. Have a beautiful day. So now we can make the whole thing a little bigger here with the font size. Let's pull a little closer. So we say a little more distance between the lines. And that looks pretty good, doesn't it? I think we can work with that. Let's turn off the AutoQueue again and do the whole thing here again. Let's see if we can get any other notes in between. No, that works with ControlM. As always, of course, excuse me. Moment. So, Control-M. Now it's back. Good. And then let's see how the whole thing behaves now. I press Start once. It starts rattling. And there we can already see how our text appears in the image generation. And that is of course extremely ugly. Let's not do anything to ourselves. That's crap. Now we can only adjust the strength with the standard note from the control net. Let's do that. So then something like that comes around. That's a little more interesting. We see here, it now starts to draw in the clouds. Down here the meadows. Our beautiful is also now in the mountains. By the way, we can only reinforce it a bit by prompting by saying snow covered mountains. That gives the whole thing a little more play space. Yes, it now has a bit of Mount Rushmore. Somehow it is not yet. And down here, that's what's happening here. Parallel dimensions. It's not really nice, is it? Of course we can turn the strength down a bit here. Yes, you can still see it a bit easily. Most of the time it gets clearer when you zoom out. But what I actually want to get out of here is that we can already see that we don't have much influence on the whole thing with the normal play control net note. Now ComfyUI itself also offers us that we can use a control net apply advanced. Here we already have a little more options. We'll change that. Now we also need our negatives here in the meantime. Clamp that in there. And that there. We hang the control net in there and our image or our text. That's our image right now. And of course that first of all behaves exactly like the other note from the strength up here. We can let it run once. Here we can already see that it is not particularly beautiful. There are a few other tweaking options that you can also do here. Here in this note, but we'll stay with this one for now. What we can also do here is we can activate timestapping. So at 10 percent it should start and stop at 90 percent. Then we can already see in the image generation that it gets better. We can look at that again. Down here with our day. But that's a pretty nice effect. When we scroll out, we can see that we can read the day. When we scroll in and zoom in, then we can only see houses that are covered with snow for some reason and the others not. No matter. With beautiful it is also the similar effect. It looks like snow-covered mountains and a desert from the zoom level here. But when we zoom out, we can see that beautiful. So that makes a big difference to the normal note that we can use here. What I just meant is that with additional points of attack here in this story, we have pure white right now. But we can also go and say, for example, we want to take a gray tone. Then the whole thing gets a little darker and also influences the picture. For example, down here we can see that day has changed. We don't have that much snow here anymore. It has become a little more green, but when you zoom out you can still see it. We can make it a little darker. That is now dark gray has now changed here. The snow has become a little less. But when we zoom out, we can still see our beautiful day. These are still tweaks that we can do here and if we put them on custom here, we can even add hex values here. 6 times f is the highest value. Let's say 6 times e. Then we get a gray tone too. So not quite as bright, but brighter than the previous one. We see it again here. There is snow added. We can still try a different seed. Maybe it's a little better with the snow or adjust the prompt. Maybe we shouldn't take snow covered mountains, but you have seen that we can already influence it here if we make the color darker in itself. Yes, everything has changed a bit again and in the direction that we of course have our font very bright. We can even make it a little darker again. But there is one thing that bothers me about this image generation or with this note and that is that we see it quite well here. We get the black background with the brightness control at this point. That's why the rest looks very dark while the font is very bright. And here comes the next game of notes or that is a note package and that is the advanced control net. We take here apply advanced control net and before we connect that, I'll show you how you can get that. Go to the computer manager, install custom notes and look for control net. And that is then here from the computer advanced control net. There is also a github page for this, of course there is still a bit of a description, you can read through the thing, because it can do a lot more, especially when it comes to the area of animations and so on. You can also deal with it a bit, try a bit. However, the documentation here at the point is not quite finished yet, if you scroll all the way down here. There are more nodes to document and show usage, we'll add this soon to do. It's really a powerful control net suite at the point, but it's not quite finished from the documentation. Nevertheless, you can already use it super well. So now let's put everything together here that the note would like to have. Let's take the positive negatives and that happens to me again and again here, if you don't hit it exactly like that. And we need the control net, we need our image and now we see a big strength of this note here, as I think. We have a mask input here and we use that directly shamelessly and say we need a little more space here so that it is a bit more tidy. We take it there. We say here now, we want to have a mask to image note, it doesn't find it like that. Then we take that out of here, mask to image. That was totally stupid, because of course we need an image to mask. It must have suggested that, right? No, it didn't. Hello, let's stay open while I type image to mask. So that's what I meant. Let's make it a little smaller and we can hang it up there now. If you're wondering, by the way, why I'm leaving the channel red here, it doesn't matter what you actually choose here except alpha, because we have gray tones or white and red is a part of white, so it works. You can also take a look at the mask here by saying mask to image. At this step, I was already a bit smaller, so and then you can take a look at it and we'll let it run again. And here we see that we have a mask that is very similar to our text. And what we can see right here on lift is that we have eliminated the dark spots in the picture and only in the masked area, which is our font, now we really have the effect we want, namely that the font is bright and rendered with it, but the rest of the picture remains untouched. And now we can of course also say again that we want to experiment a bit with the strength, then we have to touch it a bit. We want to activate the time-stepping here. I'll say 0.95 here. Let's see what kind of effects that has. Yes, that's almost too little strength. I think we'll go with 1 to the story. We can actually play around a bit with the time-stepping here, but you can already see that the pictures are much nicer. Let's take another seat, which might fit a little better now again. That's also a bit of a lucky thing. Then the A goes a bit too low for me. We'll go another step further in a moment. There is too much sky for me. That's unfortunately the case. You also have to have a seat that fits, but if you have a bit of basic settings, then it goes very well on several pictures. Maybe we'll take the clouds out here. I think he's concentrating a bit too much on the clouds. Yes, that's more of the direction. Let's take a look at the readability. I'll do one more, then I'll show you the next step. Yes, let's go on, because we can refine it a bit. When we go into the weights override here, then we also have the advanced control net suite, the scaled soft control net weights. The whole thing looks like this and I think it's actually more for animations, but it already has a good influence on the image generation. I would just go in here with a value of 0.9 at the point and we have to say flip weights. Now I can show it briefly. If we don't do it, our text in the picture will disappear more or less. You can still see it roughly in the middle, but let's be honest, we don't see anything there. But if we do flip weights now, then the whole story will be integrated very nicely in our picture. We can do the na check again. That's a bit too readable for me in the na check. But that depends on the seed. Nevertheless, we really have a very nice and smooth transition here. Let's take the time stepping back a bit. Yes, now we can of course go down here again and say we really want to have the wide here now, so that it can shine through really hard. So we still have influence options. But we can also, because we work here with a mask, we can go and say, for example, that we want to soften the mask a bit, so feathering or blurring the mask. And I'll put a blur in there, a mask blur. And at the point, wait a minute, I'll just connect it quickly and up there. So that everything is a bit ugly here, I know. The blur mask comes from the ComfyUI Essentials and you can also find them if you are looking for Essentials. Then there are the ComfyUI Essentials here. These are a small but very, very fine collection of useful notes, which, by the way, are also from, you can see it up here, Cubic. That's Matteo, the developer of the IP Adapter Plus Suite, who also picks and takes care of it over and over again. And that means over and over again, actually very regularly. And there are really a few cool notes that are very, very useful. Just as a tip, you can of course also take feathering notes or mask blur notes from other suites, depending on how you want. But the effect that we can still achieve with this is of course what we see down here. The mask is now a bit blurred, softened and that of course then ensures that we also have a much better integration into the picture itself. So now you can play around a bit with the strengths and everything until it fits you at the end. Yes, 0.8 is still readable. Maybe we'll go up to 0.85 and I'll still take a different one. Yes, it's a bit nicer. So now of course the whole thing is still a bit, yes, if we take a look inside, we open it once in the browser. That's the size, it's somehow still a bit ugly. That's why I would definitely say here we go into a second pass. That means we take a case sampler again. I press control c to copy and with control shift v I enter it again because then the connections remain at the front. Now we just hang the latent over here and the latent over there. Now we have generated a second pass. That means we send the picture into the sampling again. However, we do that with a denoise from 0.3 to 0.35. I go in with 0.3. We can also set the seat to randomize here and we take another control net or an advanced control net here from the front. I also press control c, control shift v so that everything is connected. Now we pull the positive and the negative prompt over in our second pass and reduce the strength here at the point by about half, roughly speaking, and let the whole thing run once. I think that's too big, isn't it? We adjusted the VAI. Yes, let's see why it looks so grumpy here. We have to go in with a little more denoise here. Yes, okay, a little more denoise was too little, but now the picture is already nicer than from the first sampling. I have control c here. Come on, what are you doing there? Let's take a look at the two pictures in comparison. The lower one will now be the picture from the first pass, the upper one from the second pass. And it will be a bit, it will shine a little more as you can see here above, especially in the clouds. The letters shine a little better over. I have to look for the letters right now. Here we see them again. If we scroll in here at the beautiful, if you know that this is an a, you can tell or that is a b, but otherwise I don't think so. The second pass could also be an upscaling or something, but for the convenience of it, it is of course better that we work with the same size here. When upscaling, you might have to scale up the stories again and then go in. But yes, you can do very, very cool things with it and let things shine over each other. It doesn't just work with text, it's also nice here with this fog in the background. Look, it has also gotten really nice. You can no longer see the letters here in the back. The day, if you know it, yes, comes out, but if you want to get out, you can see the letter. That's just fine with everything you can do with white tones or with masks, so yes, you can also use the whole thing for qr code creations and so on. However, qr code is a completely different story, I have also failed so far. So either you get a readable version of the qr code, but the picture is rather ugly or you get a nice picture and it is no longer readable, but theoretically you can also render qr codes with the brightness controls, because they are more or less just black and white pictures, but for texts it works perfectly fine. We can also see if we can change the prompt here corridor in space ship hi-tech intricate which I left out in the prompt by the way, intricate and something like that, of course, that also improves your picture in terms of length if you then do a bit of stylistic descriptions smoke raising we'll just take it in with high details in the letic let's see what happens yes, it works pretty well here, too, that always depends on the ziet that you take and on the description but so from the overblending really nice here I would even say almost again we take a darker gray we are up here on white we go back to custom so that everything gets a bit darker that maybe even too bright in malnit bs in yes and so you can then touch the edge to achieve really really cool wild results once again the model I think the absolute reality was a bit better for it but I think you can see the use area and how the whole thing can work we have a lot of possibilities to tweak the whole thing we can still read it now because the strength is a bit too little nonetheless it is quite interesting how you have rendered the whole thing here so you can't see it at all here it is again a bit recognizable there you could then go in with another gray tone again so we have a lot of options to make adjustments here , on the one hand, the strengths themselves, let's put a little bit on it again, maybe too much the strength we have the time stepping at the place we have the option to make the mask a bit softer there you can even take more we are now here on six on the way six pixels you can screw up and down we can give down here how much white it really should be or whether it should be gray and all of this then ultimately affects the picture which we get and that's just super a lot of settings if you first want to do what you want to do for the motif the right prompt has caught the right seed has caught and down here all the settings and all the parameters have played around a bit looked at how the picture has changed and adjusted then you can really achieve really cool results with it so pictures even after the upscaling where you zoom in close in detail does not suspect that a letter is hidden there but if you zoom out you can still find such a hidden message on the whole story and that's pretty nice, for example, the day down here that looks like light reflections the letters here well that has a bit of it has a bit of it somehow that could also be a light reflection a bit of tweaking here and there and then you can get really cool results with the target so now I have, yes, I've been talking about it for so long despite my sore throat , I still hope you have fun with this technology and can get really cool creations out of it, I wish you a lot of fun when remaking experimenting and playing around and we'll see us in the next video until then good bye
