Noise Styling is the NEXT LEVEL of AI Image Generation

this literally redefines how noise is used for AI hello my friends how are you doing today I'm going to show you something that is nothing short of revolutionary this is an idea by aat Suzi there's a workflow you can download it's more complex than mine you can also download my workflow me showing in this video to experiment with this method now I'm going to show you two things here that you can do specifically with comu ey because it allows you to inject noise I will do it in two different positions one is in the latent noise that's the base on which the image is generated and then we also going to do it in the depth map all right so let's have a look on what is happening here now we have two sections here that are pretty interesting this again gives you a very good indicator why comu ey is such a powerful tool so here we're loading an image now here we are scaling the image to 512 by 768 no matter what the input is so this is going to be cropped a little bit and then we are also adding some padding here to the side with some Feathering which means it's a soft border and then there is another upscale to make it into the right dimensions and then this is blend together to this imetry you can see this has the right Dimension that we want to have for our rendering but it also has a white border around that we're going to use that for the artistic style now up here we have the classic loading of the checkpoint we have a clip skip here we have a vae loader here and our negative and positive prompt pretty classic stuff and then up here look at that how beautiful this is we have another load image this is going into control net this is why these are just gray values in here then again we are scaling this to have the right right ratio 512 by 768 and then of course we apply the control net to our conditioning conditioning is the text prompt so that the control net is part of the image here's a preview on how that output looks for our control net input basically just so you can see how that looks now this is then going over here into case sample which is rendering and then it's decoding it into an actual pixel image from the Laten space now I did include some extra steps here so here we have a upscale so this is getting 1.5 * the size and then we have another case sampler rendering it for higher quality and then over here I'm sending this into face detailer which is basically like the confu ey version of a detailer to just give me a nicer version of the face more beautiful version of the face now as you can see here I've turned off the upscaling I've turned off the second case sample because we only need a low resolution version for me to show you what is going on here I'm also going to drag these elements over here this kind of input here and this kind of input here so remember the colorful part is the input that is injected in the noise and then over here this is used for the depth map now again also it's important to point out here for the first case sampler I'm having a d noise of 0.7 five so this is taking a little bit of the image also into consideration this is why the color here Works onto the image we're using it's kind of like a little bit of image to image rendering let's experiment with this and see where we get with this so I'm going to show you something that's more obvious so here I have painted different kinds of noise you can do whatever you want you can create any kind of structure any kind of input let me make this a little bit bigger so you see what's actually going on here this I've done with affinity photo which is a program that can do this very easily I want to show you something here is pretty cool so I have here you can see this noise that has a blue outer part and then we have this pink inner part and then when we're going to render this over here you can see this is now cropped and has this white border and it's rendering over here and it's of course also using my depth map over here and this is my output now see how interesting that is so first of all we have this kind of artistic border here which is very nice it's a combination of the white water that we have from the noise because if there is white there's no noise at all but also from this structure here that we have from our depth map that is making this interesting brush paint watercolor border here around that really really cool but you can also see that our character has red clothing and there's a blue surround Ching around that now look what happens if I take this noise here that is yellow in the middle instead so we're going to render the image again and as you can see after a short moment of waiting now our character has a blue and red outfit and the surrounding is still blue so we have an influence here on the design of the image but also on the posing of the image you can do some very dramatic interesting stuff let me show you something else so here I'm going to use a noise this one multicolored noise also has this kind of Bend in the middle so let's try this out over here let's see what this is rendering so here we have a new result and you can see how this is much more colorful but also look at the nice Dynamic pose we are getting from that that is stunning it's really interesting it's not using open pose or anything this is bringing stable diffusion to Dynamic interesting poses based on the noise that we injecting based on a depth map that we have painted by hand that is just amazing okay let's try different depth map over here so I've prepared different styles as you can see here you can really experiment with that and over time you get a feeling for what it's doing and you can actually have more control over that so for example what does happen if we have a dark blop in the middle and then some pattern around that let's load that in here you can see now it's sitting over here we have the same input we have the same seat everything is the same so here you can see the changed output everything is the same but the depth map and suddenly we have this pose where she's bending her hip to the right side standing and it's kind of like a little bit unnatural as curve but it is a very Dynamic very interesting pose nonetheless you can also Al see from The Blob we have in our depth map that we also have the background here rendered and then a lot more of this white part and then these Big Brush Strokes around the characters how beautiful is that let's try out a different pattern so this time I want to use this kind of triangle or wedge here and you can see again we're getting a different result from that where she's a little bit floating in the space we have these Big Brush Strokes in the background the legs are a little bit cut off in that space because of course we are having here in the depth map just darkness a very dark gray in there so there's not much to go on for the AI but again we have a very interesting Dynamic composition here now this bending of the character is also based on the bending we have over here in our input so we have control over what we are generating because let let's try another noise that being in checked in to our latent image so in this case for example let's try here this blue and green noise and here we have the new result again look at that very interesting pose now we have a lot more blue in there we have a little more green in there we have a very Dynamic pose with the angle coming up from down looking upwards at her so this kind of dynamic this kind of artistic expression also the playful of the background I have not seen with stable diffusion out of the box so this is really interesting and by painting that by creating these kind of noise injections yourself you've so much control over that this has so much potential for experimentation I also want to show you a little bit on how I do this inside of affinity photo that is the name of the software this is a software that is easy I than Photoshop you buy it only once you often get a lot of free updates for that and here you can just pixel paint on that so you can see here I have painted this structure that kind of looks a little bit strange down here on the right side you can see we have layer effects we also have live filters so when we click on the live filter you can see there's a long list for that and one of them is called half tone effect so what I can do with the half tone effect is the sliders here you can see I can dynamically adjust the kind of noise I want to have here I can adjust the contrast I want to have here I can do a lot of different other things to play around the color or it might be lines or any kind of other thing so there's a huge potential to create all kinds of different noise just with this halone tool now for example what I can also do here is I can do a recolor adjustment so when I click on that I have these sliders here now you can see that this is black and white so for recoloring that case what we need to do is to push up the brightness you can see this then turns it into color we can have more saturation in here and then you for example can move around this kind of slider here but of course we can also go into the half tone and set this back to color so that this is more in a color tone and then we go back to the recolor can actually go back here now we have to reduce the lightness to get more color and then again we can readjust this any kind of color just with the slider so that's pretty beautiful but you can also of course blend over that do any kind of other thing so for example here I have a hue blend mode for that then I have an hsl shift adjustment I turn that on you can see this again can switch the complete color here give it more or less saturation this is what I'm doing here so there's a lot of potential for that and then of course you can do other things like just for example I paint on that structure that I've created so I make a new pixel layer I blend it with multiply I just paint on that with different colors like let's for example say I want to use orange here so just go in here and I can paint some orange in here and I can say what do I want here let's say I want some pink in here like this so I can create all kinds of crazy patterns all kinds of different interesting noises and color combinations and of course you can also put images on that or all kinds of other things that you find online that create yourself and then you can experiment with that you can upload that into comi and of course what you can also do is use the filters because this is a photo editor so very interesting thing you can do here is to go to filters distort and then mirror and with that for example I can set the amount the number of mirrors now this is not going to show us much right now because this is round but when I click and then move it around oh look at that there's some pretty cool stuff happening here all of a sudden and just like that you can create all kinds of really crazy patterns in here and things that you then can use as an input so I would say with all of this aat Suzi really has opened up a new dimension of how to work and play with AI image generation you can download all of this right now with the link below the video let me know in the comments what you think about that thanks for watching and see you soon bye oh you're still here so uh This is the End screen there's other stuff you can watch like this or that's really cool and yeah I hope I see you soon uh leave a like if you haven't yet and well um yeah

Channel: Olivio Sarikas

Views: 41,990

Keywords: oliviosarikas, olivio sarikas, olivio tutorials, stable diffusion, stable diffusion tutorial, best ai tools, ai art, free ai tools, comfyui, comfyui tutorial, comfyui workflow, comfyui nodes, ai tools, stable diffusion ai, comfyui install, stable diffusion install, ai news, comfyui explained, ai, nodes comfyui, top ai tools

Id: mLmC-ya69u8

Length: 12min 45sec (765 seconds)

Published: Tue Dec 12 2023