It is worth mentioning that although VEAI is used for video processing, it can also be used as a single frame image for magnification purposes.
No wonder, has not yet received an engine update.
Welp, v5.4.4 is a bust for me. The compatibility mode work-around is no longer functioning so Gigapixel and my other Topaz stand-alones are back to closing immediately after being launched, faster than before even. If I didnāt know better Iād say the AI is ālearningā how to make me sad, touchĆ© skynet.
If anyone has any ideas, I will try anything - system environment is as follows:
- Windows 10 Pro 64bit version 20H2 (OS Build 19042.746)
- GTX 1080 with Game-Ready driver version 461.40 (Jan 26th 2021)
Support had mentioned a similar issue across platforms using integrated graphics, but as I use a dedicated GPU, Iām not sure what to fiddle with next.
No idea why that should not work, for what was the compatibility mode if the os is current?
Weāre still currently digging into this issue across all of our applications. In the meantime, one possible solution is to try and directly connect your monitor into your GPU - this has fixed it for quite a few people.
Otherwise, you may need to open another support ticket as we have a few other workarounds for this problem. What OS are you on?
Just curious, when will the new AI engine migrate to Gigapixel. 
Have installed Gigapixel v5.4.4.
In freestanding mode, there is a minor issue each time when an image is first opened in this version, in that the preview side of the display is left blank.
If I left click on Preview the preview will show whilst the mouse key is held down but disappears when it is released.
Simply repositioning the image slightly on the āOriginalā side causes an auto update, and the preview image then shows and remains showing.
Graphics info
Application & Version: Topaz Gigapixel AI Version 5.4.4
Operating System: Windows 10 Version 2009
Graphics Hardware: GeForce GTX 1660/PCIe/SSE2
OpenGL Driver: 3.3.0 NVIDIA 461.40
CPU RAM: 32688 MB
Video RAM: 6144 MB
Preview Limit: 7840 Pixels
2 Likes
I have the same problem as the poster above and also Iām unable to export as DNG. I click save and it just sits there as if I had never clicked it. I have a very similar system as the poster above me.
Hey Taylor,
My main monitor is directly connected, but my Cintiq display uses an adapter. Iāll give unplugging that a shot and if that doesnāt do the trick Iāll throw in a new support ticket.
Iām running Windows 10 Pro - to answer TDXās question I tried out the Windows 7 and Windows 8 compatibility settings, no joy either way.
5.4.4 install went well as expected. macOS 11.1 iMac i7. Processing images basically the same as 5.4.3. Using mainly the Compressed mode, falling back to Standard when the image seems overcorrected even when the sliders are both at zero and Remove Color Bleed and Face Refinement are off. Still good and consistent results as I am processing sets of images in batch mode. Thank you, I appreciate the hard work and progress on the application!
I am curious too. Didnāt see it released to DeNoise yet. Wondering how much Gigapixel will be delayed.
1 Like
Is there any chance you can give us a proper batch processing with gigapixel ai? I spend almost 20% of my time adding files into gigapixel ai for processing.
Why not add a batch processing mode where you specify a folder, then you specify a filename pattern, such as image0001.png to image50000.png and the program automaticly understands that it should look for next image as image0002.png, image0003.png. Then you specify which settings you want and youāre off processing.
Adding 50 000 files to your gui takes nearly 2 hours, which is a waste of time as you always want all files to have the same setting anyway.
When can we get proper batch processing and stop wasting precious time?
It sounds like you are processing a video, if that is so try Video Enhance AI.
**Known Issues
> * GPU has artifacting that CPU processing does not. A fix should be released for this in the next version.
Is this fixed already ?
1 Like
I would love to only use Video Enhance AI but the problem is Video Enhance AI sucks when it comes to quality.
The output you get from upscaling 2x mp4 CRF 15 vs what you get if you just deblock and output as png source then use gigapixel AI on the png output and upscale via gigapixel then use ffmpeg to combine the frames at crf 15 its laughable how much better quality it is and the filesize is the same.
Obviously the time it takes is increased by 20 times, it can take my computer around 5 days to process 50 minutes of video while video enhance ai only takes 5 hours but who cares ? I want the maximum quality as i donāt want to revisit this later. Video Enhance AI does not offer a solution that gives me the same quality therefor i use gigapixel ai for batch upscaling the png frames, but 20% the time is spent adding the png frames into gigapixel ai which should be something that shouldāt take any time at all.
3 Likes
Please, can you add in next version an option to generate the preview with gpu but process the image with cpu? would be very nice configure the preview mode and process mode in the options. Sometimes I need to see a quick preview with gpu but process the image with cpu mode to have the best result.
1 Like
So you want it to be āWhat You See Is Not What You Getā?
I suspect that most users here would then complain about being misled by the Preview.
5 Likes
Sometimes I need to set up noise and blur options before spend 10 minutes processing a high resolution photo with CPU. I donāt want to spend 30 minutes to get a decent result. I want an approximate fast preview before the process, even if the preview isnāt exact 100%. I only want an optional setting. If you donāt want it you can still use the current preview-process method.
You obviously will not get the best result if your preview does not match your real output.
1 Like
I agree. Somehow the concept of that takes me back to my (dark ages) Unix days when weād do a bunch of coding and have no clue what the output would look like until we ran back & forth to the printer to see the output in a series of trials (& errors!). Granted I was much thinner then for all the running back & forth and not sitting so muchā¦ but I donāt think it would help now.
