Sigma Corrected Noise

[inferno46n2]

Hello,

I was wondering if it would be possible to add a similar option to your unsampler node as the Auto1111 script has with it's "sigma corrected noise"

This substantially improves the output and looking over the code it shouldn't be that difficult to implement (hopefully).

[BlenderNeko]

There is an option to center and normalize the noise, which you can then use via the injection node. This is almost the same thing (a1111 doesn't center, only normalizes). If this isn't entirely what you're looking for I might be open to changing it so it doesn't center, if there is a good argument for it.

[inferno46n2]

Is there somewhere else I can chat with you about this in detail?

Unsure if you are in the AnimateDiff Discord etc but my socials and discord handle is AIWarper (pln1). In short, I've been experimenting with this tool (mostly in Auto1111) over the passed 2 weeks and I've nailed down a pretty incredible workflow for vid2vid style transfers that capture expression and lip sync movements near perfectly.

I'd love to chat with you about it

[inferno46n2]

Alright great talk 😂

[rzezagar]

Guys I'm really curious about results of your discussion :>

[inferno46n2]

I ended up just writing all the code myself with a friend and bypassing this node entirely

The results are now perfect