0:00 this is one of the best ways to generate a ey art today hello my friends how are you doing in this video I want to preach to you I want to convince you of something and that something is to give com F your eye a shot don't click away right now because in this video I packed a lot of reasons on why you should give it a shot one of these reasons is because comu ey is more easy on your resources so if you have an older computer this this might be the perfect solution for you and of course another reason is because it is completely portable and self-contained so it is very easy to install let's get started with this video Let's Start by a
0:45 workflow that I built for you it only uses three different nodes to create all of the text to image rendering a highis upscale and after that even an image sharpening and this is super easy to understand if you come from automatic 1111 you should feel very at home with this another benefit of this is it works very well with 1.5 models but then also with sdxl models that have the refiner baked into the model which most of the community models have the reason why this build looks so simple is because it is using custom notes built by the community and here we actually have the next benefit of comi so comi you can understand like an open sand box and
1:30 anybody out there including you can create custom notes for it now this has two benefits first of all you can create notes exactly like you need them so that they feel good to you but secondly because everybody out there in community can create these notes and the community is very active it means that new technology for AI is often adapted very quickly and much faster than with other uis for example with automatic 1111 often the UI itself and the integration of new technology is a little bit behind the state-ofthe-art and because of that extensions also have to be behind in what they can do in comu ey there is a super easy way to get these custom notes
2:17 installed and that is the manager so over here on the right side you can see I have a button that says manager I click on that and here's install custom nodes and in there I have a very long list I click on install it's installing that I can also update it can disable it can uninstall it all from this part now this itself is also an extension and it's very very easy to install that so when you go to the GitHub page for the comi manager you simply scroll down to the installation here it says you need to go to the custom noes so inside of your comu I Windows portable folder you go into the comi folder in there you go to the custom notes and then up here in the address bar you type CMD and hit
3:03 enter this starts your command window where you are already in the right location go back to the confu IM manager page and you want to copy this line here git clone and then the web address of the comu IM manager copy that into the command window hit enter and let it finish downloads afterwards you simply go into the base folder of confu I Windows portable and you click the Run Nvidia GPU bet to start com VII this will do some additional installs and maybe also downloads and after that you have the manager with that you have already the by far most important extension for comi now I'm going to give you some advises of what to download next so here are several packs that are
3:49 very helpful to you and are used in most of the other workflows you can download from the internet one is the comi impact pack the next one is the efficiency notes for com youi then there's also the W node suit and one thing that is one of my favorites is the ultimate SD upscaler which will upscale your images while tiling them at the same time so this is very efficient it also works for slower gpus for each of them you simply click on the install button on the right side and then after the install has finished you restart com UI to also do additional installs and downloads required now that you are ready to go let me show you
4:34 something that is pretty amazing here I have that workflow text to image I showed you before but here I have another workflow this is using a different model for sdxl but basically the same belt and here I have yet another workflow this is doing imageo image rendering and model switching in between which means I have two models loaded at the same time and here is another B this is from a member of my community also doing sdxl rendering but in a more advanced way now as you can see I can run all of them in different browser tabs at the same time and what you can also see especially with this here and also with that is the complete workflow everything you want to
5:23 do to the AI imag is happening inside of that canvas so there's no jumping between tabs everything is happening right now you can move everything around you can rebuild everything as you want and that is the huge benefit with this I can build the UI as I need it for my artistic project and I can experiment with different builds in different tabs of my window imagine you don't have to change any settings you don't have to go through tabs you don't have to go back and forward all the time you just switch to another Tab and then you can test out different builds different workflows different ways to go about that that's pretty amazing now if you don't know what model switching is let me show you
6:10 real quick so what I do here is I am rendering an image here with ref animated so you can see here this is a classic ref animated style looks really beautiful but I want to have this kind of composition and the colors and everything but I want to have it photo realistic so up here for my second sampler I am using for this epic realism and you can see down here when I go to this image I have the exact same composition in everything but now it is photorealistic you see even the haircut is the same how amazing is that now of course over here I'm doing my ultra upscale with the tiled upscaler and here I have my higher resolution version for that that is even sharpened because I decided to put a sharpening note down
6:55 here now here's the crazy thing all of that is saved inside of the meta data of every single image that you render the complete process of whatever you're doing no matter how complex that process is of course I hear you say automatic 1111 is also saving the metad DAT in the image with all of the steps and information and settings you did in the extensions that is true but it's only true for the last thing you did with that image so for example if you send the image from text to image to image to image and then use another model like the model switching I just did that kind of thing is not saved in the metadata but it gets even better not only is your complete workflow saved inside of every
7:43 image you can have extensive notes on what you did why you did it links to videos or resources where you got it from all of that inside of the workflow with every single image that is just mindblowing on top of that the community is Super Active about sharing their workflows and helping you understand what is going on one of the very active people here who's also part of my community is akat Suzi and she has these links here also thank you for your support for creating this video so here you have templates for SD 1.5 with Rising complexity so you can see we have here one that is a bit simpler she uses the individual notes rather than these kind of compacted notes I show you
8:29 before so you get an actual understanding of what is going on and then these are becoming more and more complex and she has a flow from the left to the right so you can go through that step by step to learn from her builds to understand how to work more efficient and creative with comi here she has the same thing these builds are for sdxl in that case again you can see here there is a simple workflow intermediate and then also Advanced with a flow from left to right so you can understand what is going on now let's talk a little bit about artistic freedom and convenience in automatic 1111 the UI is built in a fixed way you can't really change how it
9:15 works but in com UI if you think something should be different or it should run in a different order or you want to add something to the process you can simply do that you're absolutely free in how you work with it and what steps follow on other steps so here I want to show you two examples of that so before I showed you this build that goes from ref animated to Epic realism and of course the thing I did here is that I just added a second sampler into the process to just enable me to do that another thing that I told you before is that I added some image sharpening down here because I thought well the output of the image is pretty cool however I want it to be a little bit more sharper
10:00 crisper more detail detail so with the image sharpening down here I can simply inter check that and I can put that wherever I want in the process so for me I decided that the image sharpening is happening after the upscaling is done but I could just as well also sharpen the image before the upscaling is happening of course with the same logic you can add notes anywhere in the process for any kind of reason just in the way you want to do things and just in the way they give you the best results and this is the ultimate artistic freedom but also gives you full control and a lot of ways and possibilities to experiment with AI so this is a complete new universe that you
10:45 can dive into now here's another point that I want to address some people say that comu ey because of all these cables is just too complex and it's hard to understand how to use that and yes that might be true in the beginning because certainly there's a bit of a more steep learning curve for that however I can guarantee you that most of the time you will find a workflow that works best for you and then use that 80 to 90% of the time and comu eyes only as complex as you make it yourself because you buil these workflows so if you want to use a simple workflow you're absolutely free to do that and with these efficiency notes that you can see here in red and
11:30 green these combin the functionality of a lot of other notes into just two notes so again this makes things very simple at the end of the video here I want to give you a simple explanation of this workflow that I built for you I will also include the second workflow this has this model switching in it and I will talk a little bit about how that works too so let's get started with this here and with the overall idea of that UI so when you want to add new nodes to that basically what you do is you doubleclick on the empty background and you can see here that you have a search function with that you can type just anything for example vae and this will show you every note that has vae in the title so you can simply select it from
12:17 there so finding these notes is very easy there is also an alternative way where you rightclick you have here add note and there you have the different categories and packs of notes from which you can choose now let's talk about the notes that I'm using here and how they work you can see in this note you can load the checkpoint in this case ref animated the vae you can adjust the clip skip this is working negative not positive so if you have a clip skip of two don't set it to two set it to minus two you can also add a Lura here I'm not using one you have the positive and the negative prompt down here you have the resolution of the image and also the batch size of how many images you want to render so as you can see all of that is super simple another thing you can
13:03 see here is that they are connected by cables to the next node and they are also colorcoded and they also have a text here so of course model goes to model and then conditioning plus which means the positive prompt goes into positive conditioning minus goes into negative latent goes into latent image it's basically the resolution that you're using in this specific case and also the vae goes into the vae which is optional here but we are still using it next we're going to talk about the K sampler which is the render settings inside of automatic 1111 so here you have the sampler State don't worry about that in this case it's just sampling you have here the seat you can click down here to switch between minus one which is a random seat or the last seat that
13:50 has been used to generate the image then you have the amount of steps CFG scale the sampler name which you have a list here that you can choose from and then the Shuler now now the scheduler is not a thing you can choose inside of automatic 1111 but for example you certainly have he about for example using DPM Plus+ sde with a scheduler of caras now here in kuui you can select them individually experiment with them and then of course gives you a lot more choice about the different sampling methods if you want to but if you want to keep it easy keep this one for the scheduler normal and for the sampler simply use Co ancestral that often works very well with most of the models down here you have a d noise setting this is
14:38 if you use an image input over here for the latent image but usually if you go text to image you just keep this as value one the preview method is auto so you can see a little video down here every time I click on render that gives me a preview of the complete render process and then also V decode is set to True don't worry about that either next we have here the ultimate SD appscale you might know that from a script inside of automatic 1111 here's a video where you can check that out if you want to and here you can see that I'm upscaling it by two I'm using a seat number here this is also randomized again as you can see down here but you also say fixed increment decrement which means increment is counting up decrement is
15:23 counting down so we leave that at randomized you have here the steps you want to use for the upscaling the CFG scale the simpler name in this case I'm using DMP Plus+ 2s ancestral schedular caras D noising 0.25 which means how different it is from the input image 0.25 is usually a good value for the upscaling and then the rests below that you don't really need to pay too much attention to one thing you might change here is that the tile width I have set to 768 but you probably want to use in both of these values 500 112 which is used for the SD 1.5 models and then if you're going to do upscale for the sdxl models what you
16:10 want to do is to use here 1024 by 1024 all of the other settings below that you can completely ignore these are expert settings don't pay attention to them at all and then this is rendering here as tiles they are combined automatically afterwards into a high resolution image and then as a last step what I'm doing here is to sharpen the image so it's just a slide sharpening with a radius of one Sigma 0.5 and Alpha 0.3 you can either use that if you want to or you can rightclick on that and say bypass from that menu and that this will have this kind of violet overlay which means it's not going to be used you can do the same thing up here right click and
16:55 bypass and like that you will only be rendering the original image in in the small resolution of 512 by 768 without any upscaling without any sharpening this will save on GPU time until you find something you like and then you can simply put the and then you can simply put the seat up here so this is not rendered again but the upscaling is happening for you in the next queue if you want to activate these two notes again right click and click on bypass again this will activate it and the same thing down here one more thing to point out here is that the upscale model is loaded up here and last but not least let's talk a little bit about how to actually start the render process so
17:41 here on the right side you can see this area which has different functions you see here Q prompt this is the same thing as the generate button inside of automatic 1111 you can also see here that you have extra options if you click that this will give you a batch count where you can render multiple images at the same time by clicking on the Q prompt below that you also have save and load this does not save and load the image this loads and saves the workflow that we have here if you want to save the image simply right click on the image and here you have save image if you click on open image it's going to open up in a new tab now let's have a quick look see at my second build this is basically doing the exact same thing the only difference here is I have a
18:28 second case sampler and for that we're going to load a second checkpoint so here you can see this is for the second case sampler and here we're going to use epic realism of course you can use any kind of model you want to use everything else in this build is exactly the same as I showed you before and that's how easy that is now of course if you want to have more complex things like a detailer or control net I did not include that in my basic build just so you have something that is super easy to get started into let me know in the comment if you want to see more complex workflows that also include control net ad taer and a lot of other cool things that you can use leave a like if you enjoyed this video thank you very much
19:14 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 really cool and yeah I hope I see you soon uh leave a like if you haven't yet and well um yeah
