Why is my Colorize Neural Filter so slow ?

I'm having the same problem, I think I've narrowed it down to the CPU not supporting AVX2 instruction set, which was released with Intel Core 4000-series (Haswell) CPUs. I only have a Core 2000-series (Sandy Bridge), which I noticed you have as well. I've tested this on other systems with higher than Core 4000 series parts, and it works almost instantaneously.

The online tutorials are using a much smaller image. A4 isn't that huge in file size matters, but when running a complex (beta) feature like this - it could take some time. I tested on a similar sized image and it took a little over 3:20 to show results.

Could simply be the complexity of the image and/ or the server-side AI not yet having enough samples in its database to arrive faster at a solution. If no solution exists, those algorithms can chew through millions of iterations before arriving at a result. Other than that of course these could simply be the usual hardware-acceleration issues that seem to be a major disaster in 2020, so perhaps spend some time on checking your graphics hardwar, updating drivers and al lthat good stuff.

Mylenium