ComfyUI 08 Inpainting

[Music] welcome to another video on image generation with stable diffusion using comy UI in this video let's have a look how we can change the image on the left into the image on the right a completely different hairo using in painting this is the workflow that I use um here on the left we have standard uh prompt and then we have a couple of blocks to prepare a latent image based on an actual already available image and then we follow that up with a checkpoint loader and a sampler a efficient sampler of a couple of videos ago then on the right we have a a latent upscaler and an image upscaler which we can switch on once we are satisfied IED with this image okay so let's have a look what we need to prepare the latent image let's first just take these blocks out and build it up from scratch the first thing we need obviously is an image loader so let's right click add a note go to image and find the load image that's the one we need and let's put it somewhere over here well we can uh modify the sizes later the next block that we need is a VA ncoder to uh uh change this pixel image into a latent image so let's right click again add note uh latent and there we find a VA en code let's put it over there let's create the necessary connection the image is going to to the input of the encoder the vae we take that from our checkpoint and the output the latent image well we are not going to send it over there we are going to use a mask because we want to uh create a mask on our image that's the part we want to inpaint let's find that mask uh add note we go to latent then to in paint and and then set latent noise mask that's the one I like to have um well let's make it uh neat and tidy the output of it is going to my sampler and the vae and color output it goes to the input of the mask and then we have here on the image we have a mask and that is what we sent to this mask and now uh yeah we are actually prepared so let's load an image now and uh yeah well I'm going to try find that girl with a second uh I think this is her yeah this is her now we are going to create a mask which is done by right clicking on the picture and then uh open in mask editor and that this happens and I can make my cursor a bit thicker uh we don't have to be very careful it's a AI so if I take out an ear or an eyebrow that does not matter at all because uh the AI knows that this is the picture of a girl and it will automatically recreate that this is the mask when we are done we click save to note in the prompt I have written woman with short blonde hair I loaded a Proto Vision checkpoint uh and well let's uh do just 16 steps and see what comes out and if all goes well well we should get that woman that uh we saw earlier with the new hairo and this is actually all this is how easy it is to do in painting in comfy UI the workflow is available in the text that goes with this video maybe see you back in the next video and in the meantime have fun

Channel: Rudy's Hobby Channel

Views: 9,987

Keywords: AI, Stable Diffusion, ComfyUI, inpainting

Id: 7Oe0VtN0cQc

Length: 4min 38sec (278 seconds)

Published: Sun Sep 03 2023