Edit image during txt2img generation

[1ort]

Hello! I'm trying to implement a custom script that every n steps will have to overlay a copy of the image on itself. Something like this:

def apply_forced_symmetry(img_array, h, v):
            if v:
                img_fl_x = np.fliplr(img_array)
                img_array = img_array * 0.5 + img_fl_x * 0.5
            if h:
                img_fl_y = np.flipud(img_array)
                img_array = img_array * 0.5 + img_fl_y * 0.5
            return img_array

I shoveled all the code and figured out how to implement something on the step. The problem is that I can't figure out how to send the modified image array back to the pipeline and continue the steps already processing this image.

For now, I settled on the fact that I need to overload the sd_samplers.store_latent () method to call the code every n steps.

Other than that, I noticed that this part does pretty much what I want:

stable-diffusion-webui/modules/processing.py

Line 549 in e72adc9

else:

Please point me in the right direction, because right now I'm rushing from one to the other, but I don't have a clear plan of action.