ComfyUI Inpainting workflow #comfyui #controlnet #ipadapter #workflow

hi following the requests of some viewers today we will see how you can change clothes in an existing photo in this workflow which of course you will find in the description of the video you can use an existing image as a reference and with the help of the IP adapter and transfer its style and you can use text here we also have the prompt Styler that allows you to choose a certain style and a list of descriptions that I asked GPT to create which consists from colors patterns and different materials and I simply copied the list into a text input node now it really depends on what your need is if you know what design Direction you're looking for then it might be worth moving the clip text node back to the widget and simply describe what you're looking for and maybe also use an IP adapter if you have a picture for reference but if you are looking for variety and new design ideas then you should connect this whole group of nodes back to the clip text node and activate the random option here in showed text you can see the final text sent to the clip text node another thing you should see in this part of the workflow is this text box here you have to write the word that describes the object you want to replace in our case a shirt as we saw in the previous videos we use text find and replace to build the final prompt but no less important this word also defines for segment anything node Which object to look for in the image and create a mask for it the part here is here to help us make the mask more accurate here you can add certain areas to the mask if the anything segment failed to select the object accurately if you see that the mask is good you don't need to do anything in cases where there are holes in The Mask you can open the mask editor and simply draw where there are holes the role of all these nodes is to help with the Precision of the connection between the change and the original image I hope that soon it will be a little clearer but before that we need to understand that this whole workflow is actually an imp painting process we take a piece of the image exists and only changes it this is also the reason why I use differential diffusion here which should help in combining the new pixels with the existing ones down here there is a control net with depth anything quite basic here you can see the depth map it creates it's based on the image we uploaded and the result can be seen in the preview node sometimes this result will be enough but sometimes we will get some distortions at the edges and to avoid most of the distortions I use the image composite mask node to connect the original image to the new pixels we created and I use the mask we already created with the segment anything and this is another reason for having these nodes here so that we can refine The Mask we create both for the final connection and for the Imp painting stage here I have added a very useful node that many have asked about image save to which you simply paste the address of the folder where you want to save the final images and another nice thing to mention after we check that our masks work and are accurate in relation to the image we uploaded now you can choose again that everything will be random click here under Q on extra options and choose as many variations as you want in batch size now we will get different combinations and different examples according to the text we copied what is the priority of chat GPT to the random seed and IP adapter if we chose to activate it so I hope you learned and we'll meet in the next lessons you are of course welcome to subscribe to the channel ask questions and like if you liked it and most importantly have fun bye

Channel: PixelEasel

Views: 4,896

Keywords:

Id: dPVF3IwbAaU

Length: 3min 58sec (238 seconds)

Published: Sat Apr 20 2024