Gigapixel v5.3.2

HDR 32 bits input/output would be also great.

Spotted a bug in the latest update. Specifically the art algorithm.

Settings
6x
Suppress noise and Remove blur on maximum
Reduce colour bleed On.

The bug happens even with all of the above filters are turned off.

Original image:
Imgur

Curiously the bug is not present in the preview window. Only shows up on the upscaled file.

1 Like

I assume fixing the artifacting still wont solve the issue of CPU producing a less blurred image that preserves more detail?

Will the output of the different hardware options ever be consistent?

@taylor.bishop

Did it this time.

But the selected image window needs to adjustable in size.

Then it would work better, because the window to select the images is very small.

I’ve been meaning to mention that I think this is a bug. When I have preview turned off, just showing the image list, and then click on one of the images in the list to do something (such as change the output size for that image) the previrew turns itself on, making the list selection box small as you note. Off should be off.

1 Like

Spotted a bug in the latest update. Specifically the art algorithm.

Thanks for the feedback Grim! We actually have a fix for this in beta currently. If you’d like to help test, let me know and I can make you a beta tester.

1 Like

Yes, results would be so much more precise I think, especially in dark colors, where even the smallest color deviation can be spotted by the human eye.

Does Gigapixel AI, Denoise AI and Sharpen AI take advantage of RT cores in the new Nvidia RTX 30-series GPUs? If yes, how much faster is processing with RT cores compared to the GTX series? I’m currently using a GTX 1080 Ti and thinking of upgrading to RTX 3080 if there’s a decent speedup in processing speed.

1 Like

I’m interested too!!!

Until now, a 3080 had the same speed as an RTX 5000 ← based on an upgraded 2080S.

May change in Future.

The RTX 5000 is twice as fast as the Quadro P5000 (based on GTX 1080).

And Turing (RTX 20XX) as well as Ampere (RTX 30XX) get along better with Async Compute.

No, the RT cores are not used.

#Bug Report. Memory Leak after Processing 155 tiff images and stay idle after Processing for 14 hrs.

1 Like

Sequence as described under the pictures


(100% Crop) 1DX mk II - ISO 2000 - Capture One 21 + Denoise AI (2.3.6) + Gigapixel (from cut to 6000 x 4000)(5.3.2) downsized to 4K Sharpen AI (2.2.2)

Taken out of ACR via export.

who knows, but probably an easy solution would be to train a new AI model with more data or complexity than past years, and thus should provide visually better results even on GPU compared to the CPU running the current models.

5.3.2 taking for ever to download from the in-app installer!

Something I’d like to have, is instances. When I open a photo, I’d like to render many settings at once, and if don’t miss something, it’s not possible.
I mean, I open a photo, then ā€˜duplicate’ it multiple times then for each one, I apply different settings, then start batch.

Are you talking about the moral equivalent of masking and/or focus stacking?

Whole picture. I like to make many (usually 4) versions with different settings, then I combine them with masks and so on (with Gimp), to keep the best parts.

1 Like

I am having the same issue - so…

Resizing using height (which gave me a scale of 1.68 for my image) and processing resulted in (after waiting for processing bar to reach 100%) NO SAVE and immediate shutdown of the program. Processing the same image using 2X worked fine.

It sounds like the updater didn’t pull down the new model for you. You’ll need to reinstall - you can use the full installer at the top of this page and it should get you the new model.