For those interested. I do extensive blow ups on a large project I’m working on. It’s all landscapes. I need more blow up than GP allows at 6x.
If you are careful with your landscape original from raw, and get the perfect balance of not too much blur and not too sharp, then try out each model with light settings at 2X until you get a 2X that looks as natural as possible, then you can do the 6X after that and sometimes will get a nice looking result, or 1200% blow ups. Again, I feel the trick is finding the right softness level on your original file to where the GP models grab the best and make realistic looking foliage.
Of course I still have problems with blurry patches, but… when and where it does work, it looks pretty nice for that much upscaling.
Do experiment with different levels of blur. Or, if you had a jpg that was already sharp, then blur it in PS. Try Gaussian blur at 1X, then fade to 80% and then try your models at 2X.
Interesting.
I’ve only done a few over the limit of x6 Gigapixel upscales and I would have to check to see whether x10 x12 Longest edge 10000 Shortest edge 10000 all work because I cannot remember with 100% reliability what size box I used.
As far as I could see on screen the results were fine.
Somewhere I do recall seeing a Warning that greater than x6 upscale COULD (my emphasis) result in errors but it did OK when I clicked Continue or Ignore - I wasn’t blocked from having a go …
Google says:
Topaz Labs AI applications, including Gigapixel and Photo AI, have a maximum pixel dimension limit of 32,000 pixels on the longest side. This means the longest dimension of an image being processed cannot exceed 32,000 pixels, regardless of the file format. A 32,000 x 32,000 pixel image, for example, would result in a 1024 megapixel image, which is the maximum size supported.
