Map Bashing - NEW Technique for PERFECT Composition - ControlNET A1111

create beautiful compositions with ease hello my friends how are you doing in my last video I showed you a technique to use masks to create beautiful compositions with a hundred percent consistency on the masked object now in this video I want to show you a technique that is called map bashing so what kind of problem does this solve and what does map bashing actually mean now the thing here is this is using control net Maps so you're going to extract the control net maps from the images you want to use and then afterwards you build a composition out of these control net Maps now why would you go that extra step and not just use image to image by just doing photo bashing which is also something I have showed you on this channel well the benefit here is that a map has removed all of the extra data that a photo has in there for example the finer details the exact look of the different elements in the image the colors that are there the day light situation all these kind of things are in an image so when you have image to image render this will be used for the AI image but if you make a map bashing and you only have for example a head map it only has the outlines it only has the details in there but no other information so when you use that as the input for your AI creation you have a lot of variety and freedom with your prompting but still have exactly the composition that you want to create so let's get started and have a look at the process the first thing you need here is an idea for composition and I can highly advise for you to go to Pinterest and for example look for concept art or any kind of other art style you want to see and when we scroll down here you can see that we have here a lot of different designs a lot of different ideas of what could be an inspiration for your artwork now the interesting thing here is when you click on one of these images usually when you scroll down you have similar images so you can see we have here kind of a overlord God here so when we scroll down we have several of these different God compositions of this very large creature standing in the landscape and other similar designs so this can be used in a very nice way for giving you ideas but also on top of that you can use this to extract elements that you want to use in your composition for my example I'm using photos I'm searching for different elements so here I have a Greek ruin next I have a woman sitting on the ground then I also have a goose spreading her wings and last but not least I went to the side pixels to look for a pillar and I wanted to have a pillar that's photographed head-on not from below so here I'm only going to extract this part of the pillar after you've downloaded the images you might want to prepare here are some of them so here I'm in a software called Affinity photo but I can also use or photopia as a free website or any other tool you want to use in this case because we have a lot of background I'm simply going to crop the image down to a smaller size because I only want to extract the character of the image so now export that and save that to your drive in the next step we are going to go into automatic 11 11. when you scroll down you can see here we have control net version 1.1 when you click on that you will see that in the latest update you have here different quick choose options so for my choice here I'm going to use soft Edge when you click on that this is both loading the preprocessor and also the model for me take into consideration that you have a pop-down menu here for the pre-process so here you see the different methods to create that and you can see the different qualities the different results you get from that another advice I want to give you is that without Pixel Perfect you have a preprocessor resolution down here so when you set this higher for example to 1024 you are getting a higher resolution of your control net map we don't need any settings or any prompts because at this point we are only going to create a control net map so click here to load your image and to get my preview I have this little explosion icon down here so click on that and after a short while you can see that you get this map here you can right click on that save image s so here we have a bigger version of that map this is the PD net version and when we compare this to the Head map you can see that in this case for the head map we have more detail tails in the face so you need to decide for your process what is important for you next we're going to create a new file so we go here in Affinity photo to file and new I'm setting my resolution in the width to 1 536 and in the height to 1024. this resolution is higher than what I'm going to use to render the image but it also gives me a higher quality of the input map for the preprocessor first I'm going to create a black background here so I'm going to use here my rectangle tool and I'm going to drag out a rectangle over all of the background and for the fill color up here I'm going to select black then on the right side here where I have my layers I'm going to lock that layer so I can't click it by accident as a Next Step I'm going to select all of the maps I want to use in this image so click and drag this into the canvas let's drag this down to the middle and then I'm going to hide everything but the background I'm going to make the background visible again and then drag it as the lowest layer and I'm going to resize the background so it's going to fill all of my canvas like that next we have the column here you can see that there's some extra details up here that I don't want to use and and also there's a black background around that that I don't want to use so for that especially in Affinity photo you want to rasterize that layer so it becomes a pixel layer because when you import it this is starting out as an image layer that can't be edited next I'm going to take my want tool for selection I'm going to click here on these dark areas I don't want to use and then simply click on delete to remove them next I'm going to use my Eraser brush and I'm going to erase the parts up here that I don't need for the column and now I can move the color them around everywhere in the image I want and this gives me full control over my composition next I'm going to make the layer with the woman visible let's drag this atop of the column so you can see it is overlapping the column when I go over it and first I'm going to drag this a little bit higher because when we rasterize this this is rendered as the pixel resolution in the size you see it on your canvas this is specific to Affinity photo we go to right click on that layer and rasterize now it is a pixel layer and again I can use my Eraser tool hide all of the other layers so you're seeing what you're working on and the process becomes easier for you alternatively if you want to have a faster process for this what you can also do here is to use the freehand selection tool and then make a rough selection around the character so now we have the character selected you go up here to select and invert pixel selection now everything else is selected and hit delete on your keyboard I'm going to turn on the other layers again and now I can put my main character on top of the column so she is sitting here so next we're going to need the angel wings in that case I only want to use the right wing here so again I'm using my freehand selection tool or lasso tool and I'm going to make a rough selection around that Wing here like so so we don't have the background in the image don't worry too much about the other details because the AI is going to figure out the rest for you now because we want to use this Wing here I'm going to press Ctrl C Ctrl V on the keyboard to copy that part into a new layer Ctrl D to deselect and then hide the layer with the goose now you can see we have a wing layer I need two wings of course so on the keyboard press Ctrl J to duplicate hit that layer and then go up here to arrange and flip horizontally so it's going to flip on the other side now we're going to select the move tool on the top left and we're going to select both layers and drag it behind our main character now I want to resize both of the Wings at the same time but so they keep their ratio you want to press the shift key at the same time now I can move them around side to side on the characters you can also change the angle of the wings so you have very much artistic freedom and of course what you can now do select all of the layers for the character the two wings and the column together and press Ctrl t on your keyboard this will turn all of that into a group and that means now I can move all of that around to see where I want to have it or I can also resize all of that together to place it in a position I want to have let's go here to file and then export it and of course we're going to export this as a JPEG file now I'm going to go back to control net and I'm going to load my image here on the right side you see the head map we have created before don't worry this is not going to be used but because we already have a control net map as an input you can set the preprocessor to none and then you have of course to choose the model you're using here so this is a soft Edge model now here it is also Worth to point out that you can actually mix different methods together but you only should use one method in one control net map so when you have soft Edge all of the elements in here should be soft Edge because you can only select one model you're going to use in one control net but the thing is you can use multiple controllers up here so you can for example create one with soft Edge just for the main character and then a second one for example with that Polaris as the background for example for a city so you can do a lot of different things and combine them and be really creative now it is time to set our render settings here I'm using ref animated I'm using clip skip 2 and I'm using the SD vae 840 000 next let's have a look here at the prompting I'm using the prompted beautiful girl with white Wings cute face smiling blue eyes sitting on a broken Greek pillar and then with two round Brackets cross-legged Crossing her legs and after that Greek ruins and dramatic clouds in the background now below that for my settings I'm using the sampling method DPM plus plus 2sa Keras I'm using 30 sampling steps to give it a little bit more time to think about what I want to create here and here I'm using a specific trick to get give me more quality and also because the image is more complex because of the control net composition so here I'm turning on a high-res fix but I'm setting it to upscale one so it's not upscaled at all and I'm setting it to denoise strength 0.4 what this does is that it first renders the image with the given resolution and then it renders the image again based on the input with the same resolution this is kind of like image to image rendering and it just gives the AI in the second run through more information of what you want to have so you get a better output then below that I have my width at 768 my height at 512. the batch size here is one and I would suggest to you until you figured out the prompt to leave it at one and then after you've figured it out to set the batch size to 8 so you get eight variations of that image and you can choose the best one I find this is a faster and easier method to get actually good results from your images the CFG scale here is at 7 and the seat is especially if you try to figure out to get a good version is best at random so that you can see the different variations you get in between the images so here we have the render output of our eight images and when you enlarge these images you will see that they are different of course in the variations but also one thing that stands out here is that the face often doesn't look so good and I would not worry about that because this is going to be fixed in the upscaling when you found an image you like click down here on send to image to image in our image to image settings you want to scroll down here you want to set the resolution to twice the size so in that case 1536 by 1024 I'll alternatively you can also click on the resize by Tab and set this here to 2 then for the denoise setting because I want to fix the face and I want to have a higher quality of my image I'm going to set the denoise to 0.4 this is going to change the image a little bit but it's still very consistent with the original so here you can see our final result I would say it's pretty beautiful and very detailed you can also see that the face now looks very nice and here's an alternative version where I have replaced the background with a forest and here is another version where I have changed the composition also here you have a very different design where we have an Explorer walking through the mountains and as you can see here I've also rendered this in different versions so we have one that is playing out as a summer scenario and then we have a second one here with a sunset scenario for this second composition again I created a map bashing for which I used a landscape image a image of a castle and an image of a cross-country Walker let me know in the comments what amazing other methods you are using thanks for watching leave a like if you enjoyed this video 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 I wish you a good weekend

Channel: Olivio Sarikas

Views: 100,408

Keywords: affinity photo layers, affinity photo adjustment layers, affinity photo learning, oliviosarikas, olivio sarikas, olivio affinity, olivio tutorials, controlnet stable diffusion, controlnet automatic1111, controlnet inpaint, controlnet 1.1, control net tutorial, controlnet video, controlnet openpose, controlnet pose, composition, a1111 tutorial, a1111 img2img, a1111 training, a1111 webui, a1111 stable diffusion, a1111 controlnet, stable diffusion tutorial

Id: z6Xwh9G24uw

Length: 15min 31sec (931 seconds)

Published: Mon Jun 12 2023