MULTIPLE CHARACTERS In ONE IMAGE WITH CONTROLNET & LATENT COUPLE! THIS IS SO FUN!

and this video will be really really cool because today I'm going to show you how you can generate multiple different characters in different positions and in different styles using a brand new very cool extension so let's go another very cool extension that I'm going to be showing you today is called Latin couple in this extension basically allows you to separate an image in different regions in different zones with different prompts and then put them back together meaning that for example you can generate character a in the zone that you see right here then character B in the zone that you see right here and then put them back together in one signal image which makes this extension very very powerful not to install this extension it's very easy you already know how to do it you're gonna click on extensions click on available click on load from then gonna look for latent couple click install then you're also going to look for composable Laura and then click install too I'm gonna explain why later in the video Then You're Gonna click on installed check for updates and then click on apply and restart UI and now you should see a brand new section called latent couple this is what we're going to use to separate our image in different zones and to use it it's very simple so the way you determine which zone you want to create is by using these two arguments right here the visions and positions now the divisions as its name implies basically determine what is the fraction of the image that you're trying to cut for example a one by one is basically the entire image in one by two is basically only half and one by three is only a third a one by four is only a fourth etc etc and each zones are separated by a comma so as you can see right here for example we have here three zones one full image one half and the other half and the positions is where those zones are located now don't worry if you don't necessarily understand how this works I'm gonna show you a way simpler way later in the video so then you have the weights which is basically how strong the generation will be for each zone so for the first Zone it's only twenty percent and the second the third is 8 so the stronger the weight the stronger they affect and then finally you have end at this step which is basically the amount of save that it will use to generate the effect but in my personal testing you should definitely choose the same amount of steps that you have chosen right here your final image will be way better and then also you have this visualize button that if you click on it you have no imperfect vision of what the different zones are and the strength for each zone so the wetter the Zone the stronger the effect will be and of course the darker The Zone the less the effect will be applied so for example if I want to change the weight for the last zone for example something like 0.2 and I click on visualize you see now that this zone is way weaker which is why I highly recommend something like 0.8 photos over here trying to generate a new character alright so now end of Chit Chat let's actually see how this works in practice so for example if I try to generate an image with two people in it without the latent coupled extension and I want something like for example a flower garden in the background and on one side I want a beautiful woman and on the other side I want a man if I click generate for example it gives me something like this which is basically the exact same character I definitely see the man but I don't see the woman however if now we go back and we enable the latent couple extension and as you see for example we have three different zones you have basically the entire image you have a Zone on the left and the other Zone on the right and now if I click on generate now you have exactly what I'm looking for in one single image we have two completely different characters wearing completely different clothes with completely different hairstyles and in completely different positions so yeah that's really really cool and super powerful and the way to make this work is to use a very particular word in your prompt and that is the word end this is the word that you need to use to separate each Zone with its own prompt so as I said previously in this example right here we have three different zones you have this entire zone right here which is basically the entire image but also the background then you have this half Zone on the left and then you have this half Zone on the right so three different zones and if you look at the prompt right here we start by writing our prompt in the exact same order we start by writing what is the background or what is the setting of our image so in my case it is flower garden background but it could be something else like a spaceship or something like that this is where your character should be this is the location and then you're gonna separate each Zone using the word end so in this example you have for example and beautiful cute woman long blonde hair white private school uniform Blue Ribbon etc etc and again to separate the third Zone you're gonna use the same word and and in this example it's a man short brown hair black suit red tie etc etc and by using the end word in this particular syntax with the zones right here each tags in the prompt will be assigned in different region which will then put them all together and you can of course do this with even more zones more different characters and you can of course do this with any model that you want if I come here for example and I choose the anything V3 model with clip skip 2 and I click generate which gives me something like this pretty cool well weird position um but but pretty cool now of course since Nathan couple is its own extension you can of course use this in combination with control net and that's where things get really really interesting because now with control net you have even more power and control over your images so for example if I upload an open pose model this is an example of an open pose model made by a user called toy XYZ which is the exact same person that created the 3D model for blender that I showed you in my previous video and if you haven't watched it I definitely recommend that you do it's really cool and as you can see here we have three different characters in three completely different positions and you can of course create these poses inside either blender or the open pose editor again if you don't know how to do it I highly suggest that you watch the previous video on the subject so here for example we're gonna generate three different characters in three different positions using a combination of the control net open pose model and the latent coupled extension so here for example I'm going to click on enable select the open pose model no need to choose anything for the preprocessor here in the latent combo section we're gonna separate our image in four different zones you have the first Zone which is the background the setting you have the second Zone which is the region on the left you have the third Zone which is the region on the middle and you have the fourth Zone which is the region on the right and here instead of typing the numbers for each zones for the divisions the positions and the weights I'm gonna come here in the extra generation parameters and paste this line right here which when applied will automatically apply every single positions every divisions and weights for all the different regions and if you want this line you will find this in the description down below and now if I click on visualize you see now that we have four different zones you have the background the setting then the left region the middle region and the right region and then don't forget to enable the extension so then you're gonna write your prompt which again is starting by the setting the background and then separate each region with the end word and now if I click on generate and this is the final result and as you can see in one single image in one single generation we have three completely different characters in completely different positions and all of them by combining the power of control net and the lesson coupled extension which is so powerful and so cool now I know what you're gonna say well okay that's great and all but all these divisions positions weights they're kind of really complicated how exactly do they work if I want a little bit more Precision well don't worry my dear little friends because I have a solution for you and that is a little piece of software called Latin couple helper which is a very cool little tool that you can superpose on top of an image and then get the precise numbers for the divisions and positions for each region that you want and to get it it's very simple you're gonna click the link in the description down below you're gonna arrive on this page you're gonna click here and then click here to copy the code then on your computer you're gonna create a brand new folder so right click new folder rename your folder into anything you want go inside that folder click on the folder URL type CMD press enter and in the command prompt window you're gonna type git clone regular paste the address that we just copied and then press enter and then you will see a brand new folder up here and inside to run the tool you're gonna double click on the latent bubble helper.bot file and here you will see a brand new window appear and what you're going to do now with that window is that you're gonna place it over an image that you want to divide into zones or regions and then you're going to resize that window into the exact same shape and size of the image corner to corner something like this and now if you hover your mouse on the top of the tool you will see a brand new grid appear and you can now click and drag every zones that you want when you want your zones to be and you will see here all the numbers corresponding to the division and the position of the zone and now we can simply copy and paste those numbers inside the latent couple extension so let's see for example I'm going to replace this region right here and then I'm going to take this position and place it right here I'm also going to increase the wave so that we can see better and now if I click on visualize you see now that the very first region that we put isn't the exact same shape and size than the zone that we selecting right here and we can do this for every region that we want let's say I want to choose this region I'm going to take this number put it right here then take this number and replace it right here and now if I click on visualize and as you can see in the second region is in the exact same shape then the region that we selected right here and what's really cool with this tool is that you can actually make the grid even smaller and more precise and you can do that by clicking on these two buttons right here if you click on this button it will make the grids smaller horizontally and if you click on this button it will make the grid smaller vertically and the smaller green that you can have is 32 by 32 which makes it super super precise to select a particular Zone on an image and you can of course use those zones inside of your new generation for example let's say that I want something like a woman on the left a man on the right but this time I'll also want to a plane fly in the sky in the top right corner so for this I'm just gonna choose this area right here so something like this so I'm going to compute the divisions paste it right here then paste the positions so then they sample my movement to be right here again copy the divisions put it right here copy the positions and then for the final region I'm gonna put a comma for both the divisions and the positions and then I want my man to be right here so here for the weights I'm Gonna Leave 0.2 for the first Zone and then 0.8 for all the other ones and now if I click on visualize you see now that this is exactly what I was looking for the general area background then the upper right corner for the plane then the left side for the woman and the right side for the Mana and now if I write my prompt and I click generate it gives me something like this which although it's not perfect because the shape of the plane is a little weird it's still respecting the zones that I put right here and we still have the exact same characters that we wrote in our prompt which again although it's not exactly perfect it's still pretty cool now another cool thing that you can do with this extent attention is instead of generating multiple different characters you can just generate one single subject but with different characteristics for each zones so here's a very cool example let's say that I want to generate a pizza with different toppings for each zones so for example I'm gonna come here paste the generation parameters click apply click on visualize and you can see that we have now an image with four different zones each zone is basically a square quadrant and now if you write something like a pepperoni pizza and a broccoli pizza and a pizza with white sauce and a pizza with pineapple and bacon and we click generate and it gives you something like this which as you can see for each different quadrants you have a completely different topic you have a general base of pepperoni but here you have broccoli here you have a white sauce and finally you have pineapples which is exactly what we put in the prompt and the zone that we separated right here so as you can see you don't necessarily need to use the latent couple extension to generate multiple characters inside one single image you can simply generate one subject but with different characteristics for each zones and of course you can use this for faces too so for example if I come in control net and I upload an image of a face so something like this and I choose for example the canyon preprocessor and the canyon model and then decrease the weight a little bit to something like 0.3 and I'm gonna keep the same zones and regions from the pizza example so if I'm writing my prompt a beautiful cute woman with brown hair and an old woman with white hair and an older man with wrinkles and a young man and if I click on generate you see now that we have a character with a very peculiar face the first quarter is the face of a young woman then you have the face of an old grandma then the face of an old man and finally the face of a young man all of those faces combine to create a very strange character and this extension is definitely very cool because without it this kind of image would be very difficult to generate and finally do you remember that earlier in the video we also installed another extension decomposable Laura well this extension was actually created by the same person who created the latent couple extension and all this extension allows you to do is to simply be able to use Laura in combination with the latent couple extension which is really super cool when you know that this basically allows you to generate an image with completely different styles in the side one single image for example if I use the base zones of the extension and I click on visualize you have as before the same regions as the beginning of the video you have the setting in background and Zone on the left and a Zone on the right so then if I write my prompt but this time I would like to add a Lora for each one of those two characters let's say I want the woman to be in one style and I want the man to be in another one so for this I'm gonna click right here then click on Laura I'm gonna search for One Piece style this text will appear then I'm gonna look for Ghibli click here and now if I click on generate and it gives you something like this an absolutely beautiful image with two different characters in two completely different styles one using the giveaway style and one use it in one piece style and again don't forget that if you want even more control over your final image you can use it in combination with control net so for example if you input an image of characters in the position that you like and you use the open pose model and now if you click on generate it gives us something like this again absolutely beautiful image and again you see now two different characters in two different styles in the same exact position that you see in the control net panel generating inside one single image with one single prompt and that's really super cool so definitely try out this extension because now you have even more power than before I mean this just doesn't stop and there we have it folks thank you guys so much for watching don't forget to subscribe and smash the like button for the YouTube algorithm thank you also so much for my patreon supporters you guys are absolutely awesome thanks to your support that can keep making those videos so thank you so much and I'll see you guys next time bye bye

Channel: Aitrepreneur

Views: 94,908

Keywords: stable diffusion, ai art, stable diffusion ai, stable diffusion tutorial, stable diffusion art, stable diffusion installation, stable diffusion Gui, ai, stable diffusion online free, stable diffusion browser, stable diffusion beta, stable diffusion ai art, controlnet stable diffusion, controlnet inpainting, multi-controlnet, latent couple, latent couple extension, stable diffusion extensions

Id: uR89wZMXiJ8

Length: 14min 54sec (894 seconds)

Published: Fri Mar 03 2023