Seems that there is mixed experience here. Interesting that you don’t have the issue but others of us [Fotomaker] do. I’m on Windows 11 and you are on 10, but I doubt that’s behind this. Are you using the plugin? The problem doesn’t occur for me in the standalone version of GP, only the Photoshop plugin: whether I have the top-left set to original or not, I am unable to change the other three quadrants…
Hello Glenn -
Works as LR6 (desktop) plugin or stand alone. Haven’t been able to load into Affinity Photo or Luminar 4 yet, I’ll have to dig into why it doesn’t install.
When using v5.7.3 on Windows 11 it always crashes when I turn on “Face Refinement”
Application & Version: Topaz Gigapixel AI Version 5.7.3
Operating System: Windows 10 Version 2009
Graphics Hardware: NVIDIA GeForce GTX 1660 Ti/PCIe/SSE2
OpenGL Driver: 3.3.0 NVIDIA 472.84
CPU RAM: 32605 MB
Video RAM: 6144 MB
Preview Limit: 8000 Pixels
Please help me choose.
It so happened that in a short time the Windows version, the video card and its driver changed in my PC.
In addition, I accidentally deleted all the test samples of images processed by different versions of GigapixelAI, from version 5.3.0 to 5.6.1.
Now I have Windows 10 20H21 with the latest updates, GPU RTX3080Ti and Studio driver 472.84.
I noticed that after using the GPU in GAI 5.6.1, the sharpness of the processed image, especially the eyes, became very low.
I haven’t tried version 5.7.3 yet.
I am aware that image processing using the CPU allows you to get higher quality,
but before the PC configuration changes described above, processing using the GPU was mostly fine with me.
I do not know what the reason is. I’ve heard RTX30xx doesn’t have support in Topaz programs yet. Is that true?
Therefore, I ask experienced users to advise me the version of GAI that works well enough with RTX3080Ti,
or to clarify what is better: to look for the optimal combination of the currently existing versions of GAI and video drivers,
or to use only the CPU and wait for the optimization of GAI and nVidia video drivers to work with RTX30xx.
My installation saves/loads the last session layout. So, if I had two panes with the same model and closed the application, the next launch of Gigapixel loads that layout.
My Ps 2022 plugin Comparison View doesn’t do that.
I wish it did!!!
When my Comparison View launches it duplicates models I don’t want it to that are already open in another quadrant. Win 10.
GP keeps adding a sRGB profile to images when the original had no profile to begin with and “preserve original profile” is used. I would like being able to save images without adding any extra profile.
2 Likes
I had occasion today to try Adobe’s “Super Resolution” feature in Camera Raw for the first time (since I have been using Gigapixel for over 3 years I was not in a hurry…). I have read others saying Super Resolution was a “Gigapixel killer”, but I am not seeing how this is possible.
Here are a couple of results I just got on a 1-megapixel image (other results are posted here). So what are others not seeing about the shortcomings of this new Adobe feature?
This is probably because people do not check it themselves.
Many videos on Youtube, for example, are very old, but Gigapixel improves every three months.
I also keep finding that Gigapixel AI can replace the Megapixel of a camera by 100%.
Yesterday evening and today I tried again and I can use Gigapixel to upsample a 150 megapixel image downsized to 50 megapixels and it looks just as good or better than the original.
(I have to play around again and again because I can’t believe it myself.)
OK there are things that are not right but I have to look very very carefully.
In the example of a medium format photo that I downsized and upsized again, a font was not upsized correctly, but it is only 14 pixels high even in the original. ← the comparison is meaningless.
It is important not to sharpen the image before (DSLR/MSLR image) you want to resize. (because sharpening reduces the detail resolution and Gigapixel does need this little difference).
The arguments that e.g. an EOS R3 would be bad because it does not have 45 megapixels like the EOS R5 / Sony A1 are nonsense when you add Gigapixel AI.
1 Like
Thanks Thomas! It is true, Gigapixel enables me to make good use of my ancient digital and even film scan photos, whereas otherwise they would just sit on a hard drive.
2 Likes
But some say older versions are much better than new ones.
It always depends on the pictures.
I could never do much with the old versions.
It may be that they were better for smaller pictures, but many who say that the old versions were better then present the standard model which is not made for small pictures at all (in my opinion).
What is missing now is that you get proper denoising and a good RAW conversion.
So that you can work with the images from the beginning and not only after the processing when then after the enlarge problems arise.
Ok…and what about 5.6.x vs 5.7.3 version?
I can’t say anything about it, the software is getting better step by step with each update.
At the moment I’m waiting for the raw conversion to get better, so I can save myself the detour via other software and the result will be better.
Gigapixel “sees” best when it gets RAW files.
Here I threw everything at it that was available to me.
1DX III upscaled to R5 size.
PureRAW 1.5 + CaptureOne 22 + Gigapixel AI 5.7.3 + Photoshop 23.1 + Sharpen AI 3.3.5.
EF 300 mm, 4.0L, @ f/8, hand held, iTR-AF tracking, multi-field metering, underexposed, ISO: 500
1 Like
different procedure, similar result.
Please note, the upload compresses the images.
I chased the dragonfly for 10 minutes, it changed its seat several times during that time.
CaptureOne 22 + Gigapixel AI 5.7.3 (LowRes) + Photoshop 23.1 + Sharpen AI 3.3.5. (Out of Focus) + Denoise AI 3.4.2 (Severe Model)
1DX III + EF 300 mm, 4.0L, @ f/8, hand held, iTR-AF tracking, multi-field metering, underexposed.
ISO: 640,
1 Like
Wowie!
1 Like
Here is an example of “you should not sharpen before”.
I don’t know if it’s a bug or if it’s part of the denoising process, but in any case DeepPrime (PureRAW) sharpened even though it was disabled (or I didn’t read it correctly and clicked through quickly).
In any case, this process reduced the resolution and merged pixels that GigapixelAI couldn’t tell apart afterwards.
The result is an image that doesn’t look as high-resolution as it would without sharpening.
Demosaic via Pure RAW (Deep Prime) – then edited with Capture One (no Sharpening o. denoising) and upscaled with Gigapixel AI.
Converted with Capture One (no sharpening or denoising) alone then upscaled with Gigapixel AI
1 Like
Today, my 1DX III goes A1.
But you also have to mention that there are limits, at some point the A1 is superior in enlarging.
So with the 1DX III I come with Gigapixel easily to the A1 but further beyond I would have to test.
The outputfile was a 2x so it did almost 80MP and it was already a bit soft.
The A1 (50MP) should come close to the resolution of an IQ4 (150MP), at least the images I have previously downsized from 150 MP to 50 MP do.
Upscaled 1:1 and Original 1:1 (both downsized)
Capture One 22 + Gigapixel AI 5.7.3 (LowRes) + Photoshop 23.1 (just for cutting) +
Sharpen AI 3.3.5. (too Soft)
1DX III + EF 300mm 4.0L @ f/7.1
The cutout here is larger than the original image.
In case you are interested where I focused.
3 Likes
show.
What is your standard/criteria for “very bad version” and ditto for “best”?
Why do you categorize them that way? Pls explain by expanding on that comment with specifically what your concern(s) and/or likes are. Thx!
I could say, “Blue is the ugliest color.” But if I don’t provide any rationale or reasoning for why I think that, it’s just a vague opinion and there’s no way anyone can make blue a more attractive color for me… and, no way anyone can decide if they agree with me or not.
2 Likes
