Topaz isn't releasing unused memory and eventually crashes

I was processing a batch of images when I ran out of memory while saving them to storage. After crashing, I lost all the settings I had done on the images (why don’t you save some type of metadata file with each file being edited as other applications do). After restarting the program, I begin to do the work again. I watched my used memory in Task Manager and saw it going up a lot. I processed a smaller subset of the images to avoid issues again. I then closed all of the images, and all of the memory was still allocated. I opened more images and more memory was being consumed. I believe you have a serious memory leak.

All drivers are up to date.

Steps to reproduce issue:

  1. Open Task Manager and select the performance tab. Make sure you can see allocated memory.
  2. Just simply load about 10+ large RAW images. Run some settings on them (maybe just 1 and then copy to all the others).
  3. Export all of the images to disk.
  4. Close all of the images
  5. Notice the memory is not released.
  6. Load more images and watch the consumed memory continue to climb.

Topaz Photo AI [v3.0.3] on [Windows]

My computer specs:
Windows 11
Intel i9 14900k
96GB of DDR5 Memory 6400MHz.
nVidia GeForce RTX 3070Ti.

Both 3.0.2, and 3.0.3 supposedly had fixes, and improvement for this issue. What version are you seeing this on?

I can’t imagine ever running out of memory with 96GB to work with!

Version 3.0.3.


What is odd about this one is, I actually have 34GB of free memory. All I was doing too was reviewing the images I had opened and were in the reel at the bottom of the UI. I was clicking on after another to see the generated preview from AutoPilot.


After closing all images I have the following:


After closing Photo AI completely


So in the end, I have to close and restart the program routinely to clear the memory out to work on more photos.

And this batch I was testing, were just loaded and I clicked on about 10 of them to preview. I didn’t even export anything, which adds to the memory footprint as well.