0:00 [Music] let's talk about luras in AI image generation they are very important because they give you a lot more control over what you can create and also a lot more variety of the things you can do with the AI to follow my workshop you can download my workflow on open art you see here there's a download button and above that you have a green button that says Lounge workflow where you can actually run my workflow in the cloud for fre I built a simple workflow for you text to image to show you how to integrate luras we're also going to talk in this video about clip skip a little bit now first let's look at different kind of luras you can find on CI you can
0:46 find hundreds of different luras for example for different kind of dressing Styles you can find out there now what we can see here is also when you scroll down the result you get depends a lot on the model you're using depends a lot on the settings and the prompt you're using so one of the images here looks like a digital painting while another image can look like more of a photography or even a 3D style so there's a lot of variety in here and I would always suggest to you to look down here in the information about the Laura where it gives you some information how to use that so for example here we have a little bit of a text about the CFG scale and we also have a text about the weight of the Laura that is supposed to be used now here we have another Laura it's called
1:31 at more details and you can see here when we are using the Laura it is adding more details to our AI image output so it can push the model into different results giving you more or less details in your image this can be very beneficial here is another one that has more of an anime style in here pretty cool and can give you also more of a specific anime style in that case here we have for example another one from the the video game near automata and you can see also here we get some results that are fitting to that style so that is pretty cool that you can do all of that with the Laura other things you can do for example to get better skin this can
2:16 also be very helpful for models that have the style and the results you want but are not so good with the skin detail so you can apply a Laura for that to get better skin details with the same model that you like to work with and another thing you can do here is for example the likeness of people now you can also train of course the luras on your own photos to train it on your own likeness or from your friends or family things like that okay so all of that is going to be downloaded of course as a file Aura is a pretty small file and this goes into the models folder and in there into the Laura's folder so that is pretty straightforward let's go back to comu ey and have a quick look here at
3:02 our workflow now as always as in the other videos I have here basic built up so we have here our checkpoint that is being loaded and then we have here our positive and the negative prompt the empty latent the cas sampler vae decoder and the preview of the image and then down here also the vae loader so far nothing special just a normal text to image workflow so let's scroll in here and as you can see here we have a Laura here so this is a note for one specific Laura you can see in this case I can decide what kind of Lura I want to use and I can decide the strength for the model we have an output here for model and also the strength for
3:48 the clip so we have an input output of the clip also for both however you can also use notes that can load multiple luras at the same time so I'm going scroll out and we go down here here we have have a Lura stack and as you can see I can connect another Lura stack and even another Lowa stack as many as you want together now please don't overdo it with the lurus because then the AI model is getting a little bit confused in what you actually want to have but you can have a healthy amount of different luras in here let's say between one and five luras should be okay and as you can see also in comu ey the Laura is not called in the prompt the Laura is called here with these notes so in this case with
4:34 the Lura stack from CR which is comy roll you can switch this Lura the specific Lura on and off if it's used or not used and then you also have here the weight for the model and the wait for the clip so far everything is super easy here you have the choice of the lur you want to use and in this case as you can see we can load three different luras into this Noe and then of course of course here we have the output Lura stack goes into CR apply Lura stack again here we have the model and clip input model and clip output that of course then the clip output goes into our positive and negative prompt and the model of course goes into our K sampler
5:21 in here now in this case for the video I didn't put a second K sampler for upscaling so this is just the low resolution output of 512 by 700 68 but that is good enough for us to get a feel on what luras are doing before we render some sample images let's look at also the checkpoint model that we are using and the clip skip as you can see the clip skip is in between the load checkpoint here's the clip out and then here the clip goes into our Lura loader which we are not using at the moment or our apply Laura stack and then goes of course into the positive negative promt and here I can set my clip skip to
6:07 either minus1 or min-2 in this case I'm using the model ghost mix which is pretty good for anyim style images and then when we scroll down here again as you can see here we find some information about the model especially down here you can see for example it says hus fix is a must and then also here for the clip it says either to use one for the clip skip or to use two depending on what your style you want to create for example if we have another model like ref animated you want to use clip skip too and of course then there is other information down here now let's go back here to our prompt that we are using in this case I'm using a very simplified prompt so it doesn't influence the result too much I have here portrait of a beautiful woman
6:55 standing in a park cute anime style and I have ugly deformed and then embedding Bed artist embedding bad hands 5 now as you can see inside of comu ey the embeddings are called with our prompt window while the luras are loaded with a note to do our render tests I have moved the Laura stack next to our preview image and this is one of the huge benefits of comu ey you can move these notes around and arrange them in a way that is most useful to you so let's render an image that has none of these luras activated let's click here on Q prompt and there we have already the first image so let's turn on the phix dress now for this remember what this dress is looking like very beautiful
7:40 dress very cute short summer dress very very nice let's turn on phix dress and then render the next sample image and here we have our fix stress Laura now this is more of a close-up portrait but you can see the clothing style changed quite a bit we have these beautiful golden ornaments here very different style of the clothing now she even has a tiara on her head Little Flower here everything is very beautiful and you can see that this has a very different style than we have seen before and of course you can even influence that more into the direction of the Laura you're using by influencing how you write the prompt to put more information in there of what you expect from the Laura and the model to create so let's turn off the fix
8:28 stress and now we turn on down here this animation style promised Neverland and as you can see here as a result we have a different style of anime it's a little bit more cute and cozy very nice interesting dress also and looks very beautiful let's switch this over to the near anime style that we have seen before from near automata and as you can see here again we have a different style of our anime image looks a little bit different from the nose from the expression so you can actually get some very nice results with that here's another example of that and you can see how beautiful the results are now let's make a combination of the fix dress and the near anime style Laura and so here you can see we have a combination of
9:16 both of these Styles together where we have this interesting nice golden ornamented dress and then also the cozy style of the near atat Laura so let's change the prompt and go deeper into what the Lura are meant to do so in this case here the prompt is near anime style anime screen cap megumin one girl bare shoulders fix dress with golden ornaments intricate dress black cape black gloves black hair blush Cape choker collar bones dress hair between eyes hat long sleeves looking at viewer and as you can see here as a result we get something that is a lot closer to
10:01 what the Laura to do so it's always good to work together with the luras to get a different result for your image and be more specific to what the Laura actually can do now let's also try a different model on this so instead of ghost mix let's switch this by clicking here you can type ref and then ref animated let's load that I'm going to set my clip skip in this case to two and then we're going to render it again and as you can see here the result we are getting is is quite different because now we using the same luras the same prompt all of the same settings but with a different model now let's change the model again to Epic realism natural sin and see what kind of results we're getting from that and as
10:47 you can see here the result we get looks a lot more like a photo now like I said this is the base resolution 512 X 768 so the details of the skin are not great but if we do an upscale as I showed in my other video we can get much better quality and have this look like an actual photo 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
