Why would there be a difference in result quality between using the GPU vs. the CPU?
1 Like
EDIT: Upon further investigation my comments are incorrect, Gigapixel AI does display the original image. It is indeed a windows photo viewer which enhances images when zoomed in.
Can you please remove the artificial reduction in quality for the preview of the original images?
While there is doubt that Gigapixel does an excellent job at upscaling, the comparison mode between the “original” image and the upscaled image is highly misleading.
Below are three random images that I grabbed off google.
The left image is the original image viewed in Windows photo editor.
The middle image is the “original image” preview in Gixapixel.
Right is the upscaled image.
It should be pretty obvious to most people that there is a significant reduction in quality between the left image and the middle image.
1 Like
Has there been any word on when Gigapixel is getting macOS Monterey support?
There is no reduction of quality in that sense, because by enlarging the artifacts are simply enlarged.
To my eyes, the Windows previews look washed out while the gigapixel images show hard compression edges.
Which is correct in the sense that it can handle it afterwards.
I remember this was always an issue more or less because to make the preview in the output quality it would have to process the whole image which would take the same amount of time than processing the image. So to speed up the preview the quality has been lowered…
There is a thread pinned at the top of this forum, you can’t miss it:
https://community.topazlabs.com/t/contacting-support/25804/5
For most applications in which presentation quality matters, Bilinear or Bicubic scaling is used when enlarging. Gigapixel uses the Nearest Neighbor algorithm instead which is more useful for production purposes as it shows the original pixels as clearly as possible.
1 Like
Thanks for the responses @Viscerous and @TPX. Upon further investigation, I have learned that when viewing a low-resolution image, many programs (including windows photo viewer) use algorithms to “improve” image quality when upscaling the image. The image displayed in the Gigapixel AI is indeed the original image.
I will make an amendment to my original comment.
A wish for a future Gigapixel version down the line:
Possibility to upload own, relevant pictures to use as reference!
For instance: I have taken a blurry image of, say a loudspeaker, using my phone camera in poor lighting.
In Gigapixel, I upload several razor sharp images of said loudspeaker, from manufacturer’s commercial image bank, as references.
Gigapixels puts together a sharp image of my loudspeaker, using the other images as reference.
1 Like
Can you post this Gigapixel AI idea here:
The difference before sharpened or not can make quite a difference.
If you want as much resolution as possible, it’s better to give gigapixel an unsharpened image.
The output from Gigapixel is much finer when you don’t do a presharpen before upscale.
Presharpened in Capture One + Gigapixel
unsharpened before upscale Capture One + Gigapixel + Sharpen soft.
1 Like
What do you find wrt doing a DeNoise (vs SAI) before using GAI?
I have used Photolab 4 before for denoising, the data it spits out can be processed well.
But it is advantageous to let Gigapixel do the last bit of noise (if any).
If you use Denoise before, you have to fade in a lot of the original file so that Gigapixel still has enough to read.
1 Like
Hi Taylor,
I just used GAI 5.6.1 as a plugin to Ps CC 2021 (Win 10 desktop PC).
It did something I hadn’t seen before.
It’s not displaying the mode/model name in the upper left quadrant in Comparison View.
Any idea why that might be?
I tried clicking on that box then tried clicking on Std to have that operate there. No luck.
p.s. Just checked the Standalone and it’s doing the exact same thing. Doesn’t recognize that upper left quadrant as a usable view in Comparison View. And, no label there either.
I thought that was for the original?
Comparison View displays up to three different AI models simultaneously, letting you decide which one works best with your photos. We also ported a new feature from Video Enhance AI that allows you to view the same AI model in all three quadrants using varying parameters.
I think AiDon is right, the top left is the original view?!
Hi Don, That may be that in GAI we can’t adjust the top left quadrant. But it was always labeled. Just like what you’ve posted in your reply.
My comparison view - all of a sudden - has no such label (I don’t think it shows one in my snip); in both Ps plugin & standalone versions. That’s what’s puzzling. I tried to get it to display what that box is with no luck…
Any ideas why my upper left quadrant is label less? Thx.
Pls see my reply to Don above… I’m open to soln approaches! 
Have no idea why it isn’t labeled…
Odd, eh. I thought perhaps it was a flukey new design. So that’s why I questioned it. Hmmm.
Thx for the reply!
1 Like
