I've been seeing some stuff on Reddit about how difficult it is to work with ComfyUI (https://github.com/comfyanonymous/ComfyUI) and how it doesn’t work for a bunch of people ...

So I thought it must be even harder to get it working under macOS …

But about 10 minutes of work is all I needed to install it, run it, and get my first image generated. So not sure what the issue is, but it does look as if it works fine out of the box?

#ComfyUI #Python #ImageGeneration #StableDiffusion #maCOS #Ventura

@f it works a bit awkward with batches tho. You are afaik not easily able to, from a batch of let's say 4, take the third, and create a new batch with variations of that. Something I was able to do with Auto fairly easily.

I have yet to find out what exactly it varies to create those four distinct images.

@tazedhippo Since ComfyUI does not have any logic/control/input nodes (at least that I know of) you won’t be able to do selections from a batch etc. It’s just a different way of doing things to a UI like Automatic1111 …

You should however, be able to set up a new workflow which uses one of your previously generated images as the input for a new generation.

But of course, you might be looking for a single workflow which does it all?

@f Sure! I kinda did. But it took me a while to figure out such a flow. I was mainly reacting to the fact that you said it looked pretty easy to get going, but IMO that only holds true for the most basic flows. Once you want to do more advanced stuff, you need to wire things up yourself. Which I think is what people find somewhat daunting.

I prefer comfy as well. Runs less heavy here. But it is 'bring your own UI'. And that can be annoying.

@tazedhippo It *is* easy to get going out-of-the-box 😛 Or maybe that’s just me? I like the fact that common workflows can be loaded via a simple image, and the fact that it’s mostly self-explanatory once you figure out what each of the components does …

If you want custom functionality not provided by any of the existing workflows, then it gets harder because you actually have to figure it out …