[Posible Bug]

[Alexdnk]

Maybe I'm wrong, but the perform_upscale function doesn't need to be called for Fast 2x upscale (it's Lanczos 2x)? This function in this mode only increases the time at Fast 2x upscale.

Fooocus/modules/async_worker.py

Line 542 in e2f9bcb

uov_input_image = perform_upscale(uov_input_image)

        if 'upscale' in goals:
            H, W, C = uov_input_image.shape
            progressbar(async_task, 13, f'Upscaling image from {str((H, W))} ...')
            uov_input_image = perform_upscale(uov_input_image)
            print(f'Image upscaled.')

            if '1.5x' in uov_method:
                f = 1.5
            elif '2x' in uov_method:
                f = 2.0
            else:
                f = 1.0

[mashb1t]

This is intended, images are upscaled in these steps:

1. ESRGAN (upscaling x4)
2. LANCZOS (downscaling to target upscale factor)
3. (only non-fast) SD generation

Upscaling with ESRGAN first, then downscaling again provides a much clearer picture compared to just upscaling with LANCZOS.

Example

original (also SD generated)

only_lanczos

esrgan_and_lanczos

[Alexdnk]

Thanks, I got it.