Comfy ui IP AdapterV2 style transfer workflow | How to use ComfyUI for Everything with WFautomation

Video Statistics and Information

Video
Captions Word Cloud
Reddit Comments
Captions
[Music] don't forget to subscribe and hit the notification Bell so let's go so hi everyone and welcome to the MK Channel today we are going to see beautiful things like how you can get a desired result in a single word insert and how to transfer a style from one image to other images with the help of the IP adapter for this we will base our work on the previous workflow and we will also use masks and so that we can get consistent results in a uniform style in a series of images the big think in all of this is that we will show how to automate the entire process so let's go this is the previous workflow with the upgrade of Ip adapter plus you will find all links and the workflow Below in the description don't forget to subscribe to support us to test I will change the prompt for a bird in a box and like you see we have a very good result I like it so this will be the basis of our work for the final Warlow so let's take a look at that work and make some tests with [Music] it this is the final workflow the link is in the description here's the load image reference and under it you find checkpoint and then you find all the block for prompt the global input an input for the character for background and one for negative prompt at last the input for the image size though then you find the third Block it's for P adapter this will allow us to apply the style transfer one of the new features of the update then since we use 2K Samplers I put them in a block with the masks and at the end I put on Big Image output so let let Q prompt and make some tests it's tack of time this is why I speed four times and for a better understanding I put three images preview one for the image generation the second for the mask and the last for the final result we have a good output images I like it you can play with changing prop weight or images as you see the possibilities are infinite though this is the picture we got a very good picture in my opinion now I want to have a slightly less visible chest while keeping the same style atmosphere and maybe also a little melancholic as exists in the picture as you know this is why we use an IP adapter and especially the style transfer and this is what we got also a very good picture in my opinion considering the small change I make on The Prompt let's now change the prompt again and add an animal to see the effect I want to play with PRT as you know to see what IP adapter and what the style transfer can do that whyatt test with animal and human but keep in mind and according to my observation after the test the prompt on character must be simple this will avoid many surprises though let's start at the beginning and build the workflow together you can of course find the workflow in all the necessary links in the description of the video we will work with the Juggernaut XL model there is also the IP adapter set to unified loader one positive prompt a cyber buunk bird on nature and some reflection negative prompt case sampler whose various values and of course empty lat in width and height at 1024 for now it's not necessary to fix the seed so I press Q propped and this is the output let's now see what I do here it's simple I just copy the previous case sampler with the same seed and exactly the same values with vae and preview image connected to the IP adapter so that we can see the effect of the IP adapter on the final image and because I have the bird in the opposite direction I change the start at 0.25 in the IP adapter so now I click on Q Pro in the meantime let's compare the first result as you can see the IP adapter really changed the original image let's look at the color it's the most affected element that's why I said the still transfer of the IP adapter is very sensitive to color and now it's okay it's already much closer the result is really similar to the original in terms of composition but you can clearly see the effect [Music] [Music] though to prevent the undesired result we will have to use a mask I remind you that we want to automate this process there therefore we need a tool that can create a mask without drawing it manually and for that I will use batch clip egg this tool allows us to create a mask based on a keyword so in this case we'll be bird so we can connect the clip segment to the mask preview node to see the mask and because we only want effect on the background and not on the character we will apply the mask with the invert mask node and connect the inverted mask to the IP adapter so we Q prop to see what we got and now you can see the mask and we have a nice bird portrait let's change it and put a dog we have to change it in two places on the batch clip and on the text input [Music] so I think that you can have a very good result if you pay attention to the choice of a few elements like the checkpoint propped if you choose the right word and make it simple then wait in the IP adapter these are the keys of a good image generation as expected The Mask prevents us from surprises and unwanted effects and we have enough control so well in order to make this workflow automatic we can only choose the character and everything will work we just have to replace the character both in the clip segment and at the propt we will start with the propt we will search for replace and this is the node we need text find and replace so we will connect this to a text input so we search for a text input I choose this one then we will copy the text input this will be our variable so we will change the replace text to an input by clicking right and select covert into and desired input then we have to connect all that nodes we will connect everything to some show text node it doesn't matter which one then we will copy the PRT the positive one and paste it on the global Road I make some change on the prompt it's because I add the variable input I name a character X we don't need to put a negative prompt for the [Music] moment so write character X in the fine text replace node and don't forget to convert the text on input and connect it and wheel check if all connection are okay please do it like me e [Music] [Music] then let's make a test and see if everything works I choose to prompt a parrot for this test so well it seems to be okay all right let's Q prop and that's the beauty of comfy UI you can automate almost any idea work technique and build many software that will produce the desired results so I hope you learn well and we will meet in the next video you are of course welcome to ask questions subscribe to the channel and like if you liked it and most importantly have a big fun so keep in touch don't forget to subscribe and hit the notification Bell so keep in touch [Music]
Info
Channel: MunKaw
Views: 1,388
Rating: undefined out of 5
Keywords: comfyui, ipadapter, ipadapter attention mask, stable diffusion, comfyui tutorial, comfyui workflow, ip-adapter stable diffusion, comfyui nodes, comfyui easy workflow, comfyui controlnet ip adapter, comfyui ipadapter, ipadapter comfyui, ipadapter plus, ipadapter tutorial, stable diffusion comfyui ipadapter, comfyui ipadapter install, comfyui sdxl ipadapter, Subject Masking, IPAdapter Masking Animation, free ai, Best ai, IP-Adapter-FaceID, IPAdapter_plus, workflow automation
Id: ULuBhzKvWKw
Channel Id: undefined
Length: 12min 50sec (770 seconds)
Published: Wed Apr 17 2024
Related Videos
Note
Please note that this website is currently a work in progress! Lots of interesting data and statistics to come.