(KRITA AI) INTRO to Stable Diffusion for Krita PART 2

hey everybody Welcome to intelligent image today I'm going to be going over all of the features of the AI image generation plugin for CR including the live mode which uses AI to render your canvas in real time as you paint in part one I went over the installation and settings so check that out for how to get started I'm going to be making a video on workflows and how to use everything creatively and efficiently but first let's take a look and see what it can do so here within the docker itself we have a number of different modes we can choose from we can choose generate upscale live or animation under generate we have the place to choose our style here also can access our settings at any time with the gear icons here we have our place for our positive prompt and for our negative prompt we have the strength which this is the same as the denoising strength in comfy UI so at 100% the image we already have here will not influence our new image at all so if I click generate it will generate an entirely New Image but if I take this down to around 50% my option here has changed from generate to refine so it will take this image that I already have here and just change it a bit so you can see that's still very close to my original if I set this to around 75% it's changed a bit more so you can play around with that strength setting if I set this all the way back up to 100% and make a selection I've been given some new options here you have the options for fill expand which can be used to expand part of your image by using the crop tool to add an area to one side of your image and then make a selection and expand your image out that way you can add and remove content you can replace the background the replace background does require you to select your subject yourself it won't Auto detect the background and make that selection for you you also have the option for a custom generation settings here to be honest I don't see much of a difference between the fill add and remove if you click the generate custom you have different settings for the fill and the context detection and I'm assuming that the add and remove and fill all have different combinations of these settings so you may just want to play around with those and see what types of results you get I should also note that these different options are only available when your strength is set to 100% so you can see if I take that down I only have the option for refine and refine custom now so if you're not seeing those different fill options make sure your strength is set to 100% with this icon I can add a control net these allow you to control the generation of your image and various ways and I would have to make an entire video covering various uses of these but just in general you would select your control net you next have the option to select which layer the control net is referencing this icon will generate a preview of the control net so you can see here I have the line art selected and it's added a new layer showing me what that control net is seeing in this image and I can save this as a layer and use it again for new images if I wanted to do that you have here the control net strength so how much influence the control net has over your image generation and you have the control ending step ratio which I mentioned earlier in the settings this will allow you to stop using the control net at a certain point in your generation and I found that it can give a little more natural looking results because sometimes using a control net even at a lower strength it can cause stable diffusion to sort of try and overfit the image to the control net so if you set this a little lower it will stop using the control net at say 90% of the way through your image and the last 10% will be unconstrained and it will smooth things out a little bit and make things look a little bit more natural it's at least something to play around with and finally here we have the ability to create multiple images in a batch so if I set my batch size up to five it will create five images for me and I can walk away and come back when they're done if that's going going to take a while you have the option to set a fixed seed the same seed will create the same image every time if everything else remains equal so if you want to do that you can have a fixed seed and you can also create a new uh seed here by clicking this icon or increase or decrease the seed here in increments if you want to do that but normally I think you would just leave that unchecked you can specify where you want your image to be created within the queue and you can cancel active or queued Generations or all of them I should also note that when you generate an image it will create a preview here in the docker and also a preview layer here which will change with the image you select here in the docker and if you want to make that change permanent you can click apply and it will create a new layer here in your file next we have the upscale mode and here we have the option to take our current image and increase the resolution and add detail normally because each image generation takes some time you'd want to start out at a lower res resolution and once you get something you're starting to be happy with go ahead and upscale it first we have the option to select our upscale model these are models that have been specially trained to upscale an image without changing the image too much simply increase the resolution and maybe sharpen the details a bit they are numerous ones to choose from and some will be more specific to different types of images some are more photo realistic and some are specially made for anime type images you can set your scale here so how much you want your image to be upscaled and it will give you the target scale in pixels here you also have the option to refine the upscaled image and what this will do is run the upscaled image through stable diffusion again and this will allow stable diffusion to refine the image a bit more and maybe add some details that didn't show up at a lower resolution so you can select the style you want to use here you'll want to set the denoising strength pretty low because presumably if you're upscaling the image it's one that you already like and don't want it changed too much so probably 30% would be good there and once you have the settings you like you can go ahead and upscale your image next we have the live mode if I click this play button here it will generate a preview of my image that will update as I paint on my canvas this is possibly the most exciting feature of stable diffusion within CR because it allows you to create images in almost real time U depending on your Hardware so you can see as I paint here the image will start to update maybe change the strength here once I'm happy with what I've got I can click this icon and it will take my preview here and create a new layer and as you can see what's that layer is created it is now sampling that uh new layer again into the preview and it's giving me a a more refined look because I have the new seed after apply setting enabled so I can even create a new layer from that now and I've been able to change the hair color very quickly you also have the option to record your changes into a new animation layer so if I hit record and paint something you can see those changes take place and then if I I hit stop on record it's created a new animation layer for me with those frames so if I go to my settings Dockers and animation timeline I'm not sure why my frames didn't start at zero here the animation features in CR are not something I'm extremely familiar with but I think if I hold down shift and select all these I can drag them down to zero and if I go to my menu here I can set the clip into five frames which is looks like what we have and now I can have a repeat of the changes in my animation there so there may be some interesting uses for that and that takes us into the next features which is the animation and this allows you to render out either the full animation or single frames from an animation layer or a layer that contains an animation and it does require a pre-existing animation it won't create a new one for you although that would be an interesting feature in the future to incorporate something like animate diff or something like that but for now you need a pre-existing animation you have some settings here fast versus quality and I'm assuming that uses different sample steps if I select single frame it will take this Frame that I have selected here I set my denoising strength up it will change this Frame and I have the ability to select the target layer so if I select my animation layer there here so if I click generate frame it has rendered that frame in my animation layer I also have the option to do full animation so it'll render out all of the frames in the animation I'm going to try that now as you can see it's created a new animation layer for me and rendered all of those frames I have had the issue several times where it does render all the frames and provides a preview here but doesn't generate a new animation layer and I don't know why that is but this time it worked also when you render an animation it will create a folder in the directory where your file is saved and save all of the individual animation frames there if you want to go back and access individual ones later on the creative plugin is the most efficient way to integrate AI into your digital painting workflow but it's not necessary I did all of what you see here without it check out my videos on rendering digital paintings directly in comy UI [Music]

Channel: Intelligent Image

Views: 8,971

Keywords: AI, Krita, Stable Diffusion, digital painting, rendering, krita ai, krita, stable diffusion, ai art, concept art, digital art ai, painting ai, ai image generation, free ai tools, live ai tutorial, ai art tutorial, free ai learning, learn ai free, free ai image generator, free ai software, ai software, ai tools, krita install, krita guide, krita tutorial

Id: ziXTE6mC_38

Length: 10min 37sec (637 seconds)

Published: Thu May 09 2024