ACR super resolution: print test
This will be a temporary conclusion to my non scientific fast and dirty Adobe camera raw super resolution VS Topaz Gigapixel AI.
This last test is quite simple: take a 6MP landscape photo, upscale it in both ACR and Gigapixel, try to get the best settings I could, print it, scan it, compare. I also printed a reference image with simple bicubic upscale, that actually looks good, but both ACR and Gigapixel seem (very) slightly better.
Well, to be totally honest, when you just look at the prints (300pp, thermal print, custom calibrated ICC), there is really not any huge difference. It seems the long processing times with Gigapixel is not worth it. If you don't know what print is what, you just can't really tell.
A crop of the scans: ACR on the left, GP on the right. I could have chosen another crop, but it's all the same, and I was too lazy to clean the dust on the scanner glass.
A crop of the scans: ACR on the left, Photoshop bicubic upscale on the right. Not much of a difference in the end.
When you look closer, with a magnifier, there are lot of subtil differences. Indeed Gigapixel looks a little bit sharper, but not that much and there is overall a slight weird texture effect that is hard to pinpoint, A bit like oil painting. Some details seem exagerated when others are not. like vertical lines.
To me, the ACR print looks more natural and consitent but very close to the simple bicubic resize.
In the end, and logically, the differences in prints are much more subtil than on screen looking at pixel level. There is one evidence GP is either offering more details /and/or noise: the jpeg size of the exported files: GP=17MB, ACR=10MB.
Of course, my previous tests showed the results are content-dependant. But in the end, I just don't think I'm going to keep using Gigapixel: the slowness is definitely not worth it.