I read on the Internet that you can deactivate the use of AVX with Windwos. Even if CPU-Z shows that the CPU supports AVX, its use could be deactivated. Have a look here:
ALL AT YOUR OWN RISK!
Nevertheless, the developers of Topaz should react to your bug report.
I actually never had any issues with that install - for me it worked (I’ve been lucky enough never to have any issues).
Just thought I’d mention that part of the thread about 2.3.0 for reference thinking it can’t hurt knowing that the problem’s been mentioned before…
Selecting a batch of photos in Lightroom Classic (latest version)
Then Edit In - Topaz Photo AI using Original photos …
In TPAI - When I select image height in pixels on my first photo in a batch of portrait and landscape images I then click the “Select” box and then the “Edit” “Apply settings to all images” and finally “Save to Lightroom”
Occasionally it does process all selected images at the same chosen pixel height but mostly after the last version 2.3.2 it usually screws up the scaling and seems to apply the chosen pixel height as a longest edge.
Am I missing some setting or do I have to process portrait and landscape aspect images separately?
What do you mean by “at your own risk”?
Could it damage the PC? Or could something stop working?
Because I know that Vidéo AI v4 required the AVX2 to be installed and work. But I have 4.1.2 installed and it works. So why can’t I update to the latest version of Photo Ai and Gigapixel?
Go figure.
Quite a bit of work for your team. Your responsiveness to feedback and support system is one of the reasons that I continue to use and subscribe to your program. Sincere thanks. Peter
Upscale is performed after all other filters in every version of Topaz Photo AI released thus far.
Upgraded from within the app, so far so good. Changing the buttons from “Done” to “Close” is a good clarification.
I wanted to emphasise that I have not checked the tip and am not a Windows expert. As with all tips from the Internet, you should proceed with a certain amount of caution. You write that another programme that requires AVX2 works. But this is a different command set extension. Perhaps there are Windows experts in this forum who can help. Best wishes and good luck!
I’ve noticed a problem which I don’t think has been picked up on yet and have posted 6 screenshots to show the problem.
I’m using a picture from the Topaz DeNoise AI purchase page as I think it shows the problem distinctly. The photo was taken by Brian Matiash and I complement him on his work and hope he doesn’t mind my using it. If it helps to the betterment of the software then alls good.
Picture 01 Original picture
Picture 02 Topaz selects REMOVE NOISE, AI Model NORMAL. Result is not very good.
Picture 03 I select STRONG with the same settings and there is little difference.
Picture 04 I select EXTREME and there is no difference
Picture 05 I select EXTREME and reduce the STRENGTH by 1% and look at the difference.
Picture 06 I go back and select NORMAL with no changes to previous and its the same as EXTREME.
If I then reduce the STRENGTH on NORMAL by -1% it reverts back to Picture 02.
I think there is a lag when you change AI model which is only over ridden when you alter the strength on the new model. The erotic results don’t inspire confidence though.
I hope this makes sense. Regards Bob.
By the by I still don’t like this interface and 2.4.1 is slow and it crashes a lot.
Note if you hover your mouse over the thumbnails of the pictures I’ve posted it displays the picture settings as I described in my description.
That’s great news! It was an odd sensation when it rippled or buckled.
I’ve been busy this week. But am hesitant to update given several factors… I’m still back at 2.3.2.
1- Not knowing the current state of the right settings panel, 2- ditto any processing status box still being overlaid on images (in 2.4.0 it was centered over the bottom of my images, which makes that interface design unusable for me & my workflow) & 3- I’ve seen posts with snips here in rel. 2.4.1 discussions showing devastation to originally readable text (so not only did Text Preserve not work for them, but there wasn’t any improvement/recovery either like I’d come to expect); I use that Text function fairly regularly to compensate for the effects the other tools can impose on my text elements.
These are things I’d explore/test in a beta release. But I can’t afford to keep going back & forth (installs in recent commercial versions have run for around an hour) just to test. I need a reliable version for my actual projects vs testing (I only have 1 PC & 1 monitor).
Thanks.
I’ll wait for the next update then.
Or maybe just stay with 2.3.2 until a new major version.
I honestly don’t know how people can work with this 2.5 
I guess just working with 2-3 pictures is fine. And the frustration stays low lol
Why does “Upscale” take forever to finish when applying the enhancement and when saving the file?
I wouldn’t want to “upscale” a noisy out of focus image.
I’ve noticed something in TPAI v2.4 - did the generative part get a cut in options accessibility?
It’s impossible to select the ‘reach’ of the mask, or the way the tool processes (with clearing the masked image or leaving the image underneath intact) anymore. I was usually putting one-pixel masks outside the filling point, to have the AI more information about the surroundings.
Also, the ‘Undo’ button clears the last drawn mask as well, which is REALLY annoying - I’d just like to retry the fill, until I find the result I like, but I don’t want to redraw the mask again and again… 
Could you also put the ‘speedy’ zoom-in with a mouse roll back? It was so convenient to quickly do that.
It doesn’t make sense to sharpen a 300x400 images, and upscale the sharpened low resolution image.
That’s my point. The software doesn’t work like 2.3.2 anymore.
Upscaling at the very end doesn’t make sense.
It can only work for already good quality image.
IF we could choose to upscale first, like before, and add the other enhancement after, I’ll have zero problem.
Besides that, the text below the upscaling clearly talks only about speed, not quality of the final result. It’s at the end just because it’s faster.
I mostly work with very low resolution image. Denoising and sharpening was always very good because it was applied after the upscale.
Being able to rearrange the enhancements in the stack would remove all the frictions added to this release - I never comment on any products online. But this update broke a tool I was using almost every day. The results are 100% a regression for my type of work.
I was mostly using the CLI, but now since Gigapixel can load thousands of files without choking, I switched to that. But it’s the what Photo AI did good, the sharpening/debluring after the upscale
Interesting, some people seem to indicate up-scaling was always last in PAI.
Pls. See post #44 above from Anthony. That’s what he says there.
I’m just interested in your experience on this. Not contesting your comment …
Do you see differences in quality if you denoise & sharpen 1st then upscale the cleaned/improved image?
Does the upscale in some way alter what you processed vs just making it larger? If yes, what are you seeing change?
Or, do you just like working on a larger image to do your processing?
If the upscaling carries your processing (denoising, sharpening & whatever other funcs. you might use in PAI) to a larger version does it still matter if the upscaling (or downscaling) is last in the processing?
