ComfyUI Layer Diffuse transparent image workflow
HTML-код
- Опубликовано: 31 окт 2024
- Comfy-UI transparent images workflow
In this video we will see how you can create images in comfy-ui with a transparent background using layer diffuse, which opens up many possibilities for different combinations, compositions and animations for us creators
#comfyui #stablediffusion #transparent #layerdiffuse #composition
follow me @ / pixeleasel
workflow
drive.google.c...
Layer Diffuse GitHub
github.com/huc...
Great tutorial, thanks! It's weird though when you transform the image, you don't transform the mask, even though the transform is done after the image generated... but it works, still kind of weird, i would have assume you need to transform the mask too
thanks for the comment! when you use the decode (RGBA) it's including the mask in the alpha channel
You can not prompt anything you want with this work flow. How would you make it so you prompt it with a missing bg and then use it with videos? like a vid2vid workflow?
I found your clip a month after your first posting, and I have a problem install Comfy Layer Diffuse. There is an error in the Comfy Manager warning about the custom nodes conflicting with several nodes in ComfyUI-Inference-Core-Node.
I'm having the same problem, did you solve it?
@@UmrealEngine I haven't found any solution yet.🥲
I don't remember having a problem with the installation. What message did you get?
@@PixelEasel I think it is an issue with the diffuser version compatible with the Layerdiffusion Custom node. The custom node requires 0.25.0 but my version is 0.28.0.dev0. In the GitHub of the custom node, the developer said it is a diffuser issue, so it needs to be filed in the diffuser repo instead. 🥵
Thanks for the tutorial, when I tried duplicate it, I found after layer diffuse decode, my image is a bit overburned. Any clue why it is?
Not sure, but these models have their limitations, try reducing the cfg
@@PixelEasel I tried, and no changes. But when I use Conv, it seems fine, and when I turn Attention injection with a weight of 2.0, it's fine. Anyway, issue solved. Thanks!
I have had trouble using the Layer Diffusion Decode node on a picture I created with Ultimate SD Upscale. Any pointers? As you know, the input and output of Ultimate SD Upscale are images and not latents.
I figured it out
good to hear!
@@Artishtic could you share the workflow man? im struggling with this
At 0:58 you say you'll use "other tools which we'll see later". What "other tools"?
The intention was, instead of using the tools that are in this package, such as the tool for combining the object with a new background, I prefer to use other tools, here I chose to work with the image
composite masked node
Conflicted Nodes:
LayeredDiffusionApply [ComfyUI-Inference-Core-Nodes], LayeredDiffusionCondApply [ComfyUI-Inference-Core-Nodes], LayeredDiffusionCondJointApply [ComfyUI-Inference-Core-Nodes], LayeredDiffusionDecode [ComfyUI-Inference-Core-Nodes], LayeredDiffusionDecodeRGBA [ComfyUI-Inference-Core-Nodes], LayeredDiffusionDecodeSplit [ComfyUI-Inference-Core-Nodes], LayeredDiffusionDiffApply [ComfyUI-Inference-Core-Nodes], LayeredDiffusionJointApply [ComfyUI-Inference-Core-Nodes]
did you update comfy?
Same issue, comy update messed me up.
is it anyhow possible to upscale images?
good question! need to check...
Your tutorials are great. But if you're so good at AI graphics, why does your robot voice sound like it's from 1970s? Can't you apply your AI skills to use a much better AI voice (ElevenLabs etc)?
Thanks for sharing your thoughts! I'm all about improving, Appreciate!
*Na saaaar your voice is very very goodie *
im so tired with all this ai voice generated video, im disliking all of them instantly
sorry to hear... but u really don't want to hear my voice
Some aren't native English speakers and I find this vastly better than strong accents
@@3k3k3 true