Transparent AI Images in ComfyUI

Video Statistics and Information

Video
Captions Word Cloud
Reddit Comments
Captions
last week I created a video about how you can generate images with transparent backgrounds like this in stable diffusion by using an awesome new plugin named layer diffusion and one of our viewers requested a tutorial of the Same by using it in comy UI it is actually pretty easy to get this in comy UI you just need to download the workflow from this GitHub page this kind member of the community created nice little workflows that can incorporate the layer diffusion plugin into confy UI you can download the workflows just by clicking on this buttons here there are several different workflows of different aspects of layer diffusion you can generate images with completely transparent backgrounds you can create images with transparent backgrounds for a given specific background then separate it from the background give an image with transparent background and create background for this one and also separate that from the original image there are so many little and outstanding things in this particular plugin when you click on any of these buttons it opens another page just click this button here and download the workflow we are going to Lo this workflow in our comy UI click on load and double click on this work you can see one of the nodes in this workflow is red out for me the reason is I haven't yet installed this particular node in my confi just go to confi manager install missing custom nodes it would automatically fetch the missing one and install that once done just restart the confy UI now I'm going to show you a few different workflows for different aspects of this extension I'm first going with this first one generate foreground which is technically image with transparent background I'll be selecting jaggernaut Excel version 6 here and going with a basic and simple prompt and apple high quality the first time you run any of these workflows it would take a little bit of time because it would download required files from the original layer diffusion repository and we get this beautiful image with completely transparent background here the image is showing inside a gray out background but after using layer diffus you can see the background is transparent for the image let's test it out I'm going to download the image and place it on top of this canva page to show you how the background is actually transparent you can change the background and there is nothing and if you focus on this minute detail it has even gotten the shadow of the Apple in here all that work in Photoshop going into removing the background of an image then adding a custom Shadow to it everything is done in mere seconds let me show you something even more interesting change the prompt to a glass bottle every time I see how marvelous this is I'm simply in off what it can really produce let me show you what I mean just download this image and placing it on a background shows that this middle parts of the bottle this is not gray this is a semi-transparent bottle this is beyond our imagination if you change the background color this specific portion also changes color this is nearly impossible to achieve cutting out a bottle from a photo in Photoshop and the only viable way to get this kind of detail is by painting it up from scratch which requires massive skills but this extension does this in mere second seconds now going to push the boundaries even further by trying to create something like a person with Messy hairs and this certainly delivers there are some issues with the eyes we have to work on that on the prompt but by creating an image of a person like this you can easily use that image in any poster Banner anywhere without going through any hustle to remove the background and cut out the hairs precisely in Photoshop let us try some other workflows we are going to create a foreground for this given background The Prompt was old man sitting high quality although it has created The Old Man and made him Sat On The Bench correctly there it has also added another chair let's try running it again and this time it looks quite better what really surprises me that it analyzes the image and realizes that there is a bench there and where the old man should be seated that is something completely Beyond even what we had just one year ago and this extension is just in its early stages right now now we are going to do something even more interesting now let's try creating background for an image which has transparent background we are going to get rid of this image resize part because we don't need to resize the foreground I'm going to use this image of this dog dog sitting inside a room and this has turned out pretty well let me change up the prompt a bit changing it from the room to the jungle and this one also works out pretty well there are still some limitations of the plug-in and that is to be expected given it was released just a few weeks ago but this is a major leap towards future and whatever I showed is the list of it there are many other features of this plug-in you can pick out the foreground image with transparent background once you have created it for a specific background you can add another character to that image then pick up those characters even more specific way create the background for an image and then change that foreground to something else Marge to different images there are so many things that you can do that I can't fit in a tiny video just play around with it and showcase to others what you are doing if you would like me to create a Discord Channel where you can showcase your images just let me know immediately in my opinion confi does not do justice to this extension because you have to keep changing workflows and there are many limitations to that it Wass based with stable Forge for which it was initially created and if you want to run it through stable Forge the video is right here
Info
Channel: Debargha Bhattacharjee
Views: 320
Rating: undefined out of 5
Keywords: generative ai, a.i. generated art, Stable diffusion, stable diffusion tutorial, ai, generative art, ai image, comfyui, layerdiffusion, transparent ai image, artificial intelligence
Id: TQuU1W7ZLpQ
Channel Id: undefined
Length: 5min 12sec (312 seconds)
Published: Wed Mar 27 2024
Related Videos
Note
Please note that this website is currently a work in progress! Lots of interesting data and statistics to come.