haoming02 / comfyui-diffusion-cg Goto Github PK

Hello. Tried the LCM workflow and seems the ksampler provided doesn't work with that?

I imagine there is some work / tweaking that needs to be made to use this with Stable Cascade or am I missing something in how I'd use this node with the different stages (a,b,c).

as this plugin is easier to connect to image generation, i found quite hard to connect in processing img2img as node Attach between the empty latent and KSamplernote space to attach image

Hi - do we need to use this with the updated Hook Recenter XL node? It seems to funk up the output a bit. Im running the turbo xl / custom sampler combo (sorry) with the toggle on on the Recenter XL node. It seems like its not needed?

Hello 👋

Your research here has saved me an absolute headache and a half recently. I've found myself depending greatly on the Recenter node to clear out that richly coloured haze from some checkpoints.

The issue I've noticed is that you can completely swap workflows, and forget you had something enabled, and it still applies, despite being a totally different workflow.

Obviously, the solution would be to make the effect less-than-global, although that's much easier said than done.

Sadly I'm no richly eloquated python individual, and I'm still wrapping my head around ComfyUI's implementation. As a result, I'm turning to you for this question.

Cheers!

just wondering if you have explored high colordepth output from SD? From what i've seen we can save out only 8bit decodes

Hi thank you for the great work and article. Trying this out with SDXL Turbo and its working fine but without SD Turbo scheduler support its worse than expected.

sd turbo workflow is to use these nodes