Composition in ComfyUi using IPAdapter and Image Preparation
Vložit
- čas přidán 9. 04. 2024
- A bit of a rambling video but I wanted to get across the bare bones of composition as well as how to exploit the image nodes in ComfyUI to arrange compositions. I'll post a workflow once I have broken it into more modules to make it clearer.
Here is a Video going into the first modules in more detail.
• Composition Modules in...
and here is the workflow, which is now broken into easier-to-understand modules
drive.google.com/file/d/1cH6X...
Hi Rob. Thank You. This is fabuless. It really helps a lot.
Thanks Rob to share such great things to us. Greetings.
You are very welcome, I hope they help!
Thank You
Awesome
Always looking forward to your videos. You got some pretty nice results in this one!
Have you tried the remove background nodes? You can probably isolate subjects from images within comfyui, so you don't even have to switch to Photoshop at all in lost cases. It has transparency options too.
I have but they only really work on quite well-defined areas or objects, if you want a bit of forest or something then you are into guiding the SAM detector by hand or mask editor.
I love your videos and they are very helpful to true artists. Just one thing I wished you would do differently, and that is to not move around the ComfyUI image so often and rapidly. It is rather disconcerting. You probably don't even notice you are doing it, as you are talking about more important things at the time, but it makes it hard to see the ComfyUI layout and is even distracting to what you are talking about. Just drop the stylus when you want to talk. Then it would be perfect!
Noted useful feedback!
good one. are you dropping the workflow? thanks
Yes it needs tidying up so as not to confuse, later today.
@@robadams2451 Please also add links to lore. Otherwise many people simply won’t find them.
What does "burned" mean in the context of AI art? I haven't heard it explained. Is it "too much contrast"?
It is when an image is over processed, it leaves a slight mottle of residual noise. If you make an image to image without changing the seed you get more severe effects such as crude contrasts. It also occurs when the Model has failed to find a complete solution to denoising an image. This can be cured by making your prompt more clear.