Getting the noise from a seed as an image

[ghost]

Is there a way to just generate the noise that is specific for a given seed and then save it as an image or in an object for processing?

I know about "shared.state.current_image" but grabbing this after the first step means the sampler already has done the first step.

Also, is it possible to "inject" my own noise image and overwrite whatever the seed would produce as noise? What class or class variables would need to be manipulated by lets say a script

I am not a very experienced python coder and trying to find this out myself by looking at code would probably fail or take ages

[AugmentedRealityCat]

On the wiki there is a sample script that allows you to save all the steps that leads you to a result. Normally that should give you what you seek.

As for injecting your own noise, that's exactly what img2img is doing in fact. The basic difference between txt2img and img2img is that for txt the noise is generated by the code, while for img2img instead of noise you use a picture. That picture can be anything, including noise.
There was a nice example on reddit yesterday showing how you could apply blur on noise before generating an image from it with img2img and the blurred areas would appear out-of-focus on the final image as well.

[ghost]

Interesting, thank you. I would have to use img2img with zero denoise and then put my noise sample into it. Why didnt I think of this earlier...

The script that lets me save all the steps doesnt save Step Zero, the pure noise. Which is what I want to grab and modify with the noise of another seed. Basically what the "Seed Travel" script is doing but I want to finetune it even further. Also, if I inject my own manipulated noise picture with shared.state.current_image, does the sampler use this to continue the sampling process or is shared.state.current_image just a copy of what the sampler is doing?