Please make a design with rounded corners and transparent surfaces and such. The previous versions looked much better and in my opinion also clearer. Except for the new icons and the app icon, the design has rather deteriorated
Is there a reason why Topaz needs to redownload Models EVERY time I re-use it? I’ve reported this bug in 2.4.2 and Beta 3.0. Am I the only one experiencing this? Or is this the new normal?
Reach out to the support.
If it would be the NEW NORMAL the forum would be FULL of this info.
An odd thing: the “standardized Topaz UI” is not standardized across all apps, leading to some useability issues.
Photo AI and Gigapixel AI can no longer be snapped into convenient layouts by window managers such as Windows PowerToys FancyZones, making them tedious to manually adjust to manageable sizes on the fly to fit changing needs of workflow. Video AI can be snapped into any preset size and position with a quick shift+drag motion as normal.
Photo AI and Gigapixel AI cannot be resized when first opened, grabbing and dragging the edges of the windows fails for multiple attempts and then, finally, the mouse can grab the edge and resizing becomes possible. It is not clear what random mouse action in the attempts to grab the edges makes the windows finally become resizable. Quite frustrating.
Photo AI and Gigapixel AI have square corners, Video AI has round corners.
So, there is something different about Photo AI and Gigapixel AI where their windows are no longer fully functional on Windows 11, while Video AI remains fully functional.
Incidentally, Video AI only has one embedded icon size, which means Windows must rescale it on the fly, Photo AI and Gigapixel AI have 6 icon sizes (a subset of 5 of the standard 8, plus one extra). The colors are uncomfortable on wide gamut monitors and the style stands out as not matching typical application icons of the Windows 11 era. The black background only works with light mode desktop, not dark (admittedly, I always use light).
The reason why I noticed the latter point is because I have extracted the icons to see if I can tone them down a little (de-saturate a little for wide gamut mode, maybe add a little 3D glow highlight or simply lower the contrast a little to soften them). At the moment they stand out like a sore thumb when pinned to the Start Menu.
If you are using TPAI V3 as plugin in LrC via “plug-in-extras” and expect to get DNG as output, the output now is a TIFF. I think it apples to Capture One and … too
For details
Since v2 was introduced, I’ve found that v1 works better on most pictures, but sometimes v2 is better, and for some either one works well. This ist rue for both standard and high-res. It is nice to have the choice.
I eagerly installed version 3.0 hoping to experience substantial improvements and was immediately disappointed with the result.
Without manual adjustments and only with autopilot, many artifacts in the bright areas, disconcerting transformation from NEF to TIFF without information about the sizes of the original and resulting file appearing, as before. And worst of all, the result completely changes the color compared to the original image.
Very disappointing and frustrating… unless I’m making some mistake that I haven’t been able to see.
This issue is similar to what I reported on Nikon Z8 - the camera matrix that TPAI is using for this camera is incorrect.
Mine is Z8 too, but until 3.0 I did not found any remarcable isue. Aniway it woul be nice if Z8 is well configured in TPAI
I would like to know this as well…
I love the flexibility added to Photo AI recently. Appreciate it
Until the memory leak is fixed (introduced in 2.4.x) one can not run batches of much more than 100 images (about 120 seems the practical limit) and by no means “thousands”.
Until the memory leak issue is fixed, one can only process about 120 images before the memory use grows beyond 15GB and the processing time becomes unusably slow.
I have been using or trying to use Photo AI for video frame enhancement, however since 2.4.x it is not practical to do.
It has not been fixed in this release.
Thanks for bringing this to my attention. I’ll avoid that frustration for now.
My system has 256GB of RAM and so I can process around 120 images before the program either crashes or the processing time becomes unreasonably slow. At about 120 images the memory leak causes Photo AI to consume in excess of 17GB of RAM.
I let Topaz Labs know about the memory leak issue in 2.4.x on March 12th, 2024. Still no fix…
This is the last reply I received on the matter. I first reported the memory leak issue in 2.4.x on March 12th, 224.
From: Topaz Labs Support help@topazlabs.com
Sent: Thursday, April 4, 2024 11:37 AM
To: Jay Hightman
Subject: Re: Apparent memory leak in Photo AI
Hey Jay,
We are still working on it. The backend team has a lot on their plate and has not gotten around to fixing it yet.
–
Lingyu Kong
Product Specialist
Topaz Labs
At best, I was able to batch process 120 images before the memory leak causes Photo AI to use in excess of 17GB of RAM, the processing time of each images increases as the memory use grows.
My system has 256GB of RAM, for those with much less RAM in their computer, the total number of images able to be process is certainly significantly lower.
These reports make me wonder if there is an issue primarily with Nikon raw images. Mine is Olympus and seems to work well on Auto Pilot. I have some files from Sony, Panasonic, or Cannon, but have not tried those yet.
…Sam