Topaz Photo AI v1.1.0

There is no progress indicator when processing multiple images. Also, the batch processing seems to be very slow (on existing DNG raw files). Switching to “Apple M1 Max” instead of “Auto” does seem to have a marginal improvement, but, the average CPU utilization is around 100% instead of something like 800-900% as I get with Lightroom when I build previews or convert my NEF files to DNG.

Basically, the app is not utilizing the Apple M1 Max and the “Auto” selection for processing is selecting the wrong mode.

Hi Nitinchandra,

How did you find out that TPAI is selecting the wrong mode when “Auto” is selected? How fast is your processing when you set “M1Max” compared to “Auto”?

I was having the same problem. I ended up un-installing 1.1.0 and downloading 1.0.10 and installing it. The face selection is working ok in 1.0.10.

Here is a screenshot of what I’m seeing. It’s like the faces are drawn in at 50% scale on top of the rest of the preview. When exported, the file is saved properly, but the preview being this screwed up makes it impossible to understand what the final file will look like in advance.

This is using the latest update as of this morning - 1.1.0

1 Like

Re-raising the issue of crashing: this is extremely frustrating to spend 20 minutes going through 30 old JPEG images, tweaking things here and there, then doing an export and it crashing after a few seconds.

This would be less of a problem if it would save the user-defined settings per image, but of course it doesn’t. Has it been this unstable all along, or did I just happen to buy it at it’s most unstable? :confused: I’ll have to start loading in one image at a time, exporting, then doing another in the hopes I can avoid wasting more time.

It’s really an amazing program technically, so this is an unfortunate problem that’s sapping my joy of using the program.

1 Like

I just tried the other modes and monitored the cpu/gpu usage in the activity monitor. Setting to M1 Max seems to use more of both and a rough estimated time of processing 10-15 files showed that it was marginally faster and therefore the conclusion that the Auto mode is selecting the incorrect instruction set…

And here is another example that TPAI’s Denoise algorithm is useless on RAW images and destroys the entire image.

Here the Canon CR2-RAW, 100% view, shadows and dark areas extremely lightened, so that the drawing loss in the dark areas becomes clear. No more details in the trees. Only a uniform gray area. This makes the image unusable.

And here is the same image previously converted to a TIFF. Here you can see very nicely all the details in the trees.

The Denoise function in TPAI on RAW images is buggy and makes the image unusable.

1 Like

Hello, as of v1.1.0 I have lost GPU acceleration on my 1080Ti when using Sharpening with Lens Blur / Motion Blur models. Standard Sharpening model is GPU accelerated, but Lens Blur / Motion Blur will fall back to the CPU although it’s a lot slower (even if it’s a 5950x). Processor is specifically set to GPU, I have tried on Auto as well. vRAM is not used up, 6-7GB free when starting Topaz.

Any chance for a fix please? I constantly use sharpen on Motion Blur and it’s a big slowdown.
On the other hand, the standard model has improved considerably in a short time.

Hi Nitinchandra,

thank you for your feedback. I cannot verify the behavior you describe. I have run several tests now.

Platform: MacOS Ventura, Apple Silicon M1Max, 32 GB Unified Memory
Source image: TIFF, size 150 Mb
Tool: TPAI v1.1.0
Output image jpeg, quality 1%, size 500 kb

Auto: CPU in peak up to almost 100%, GPU in peak up to 50%, time 15.1 sec.
CPU: CPU in peak up to 70%, GPU in peak up to 40%, time 31 seconds
M1Max: CPU and GPU similar to setting “CPU”, time 31 seconds

The “Auto” setting leads to the fastest processing and highest utilization of CPU and GPU.

And here again a bug of TPAI 1.1.0:

Opened as stand-alone, no image loaded, 100% CPU usage, it never decrease. Duration until I could load an image at all, about 25 seconds!!!

And here I dragged an image directly onto the TPAI icon and opened the application with it. CPU usage: 0%. Duration, until I could use the TPAI about 3 seconds!!!

Please Topaz engineers, there is something wrong with the code. TPAI as stand-alone application needs 25 seconds until I can open an image at all. Also, the idle CPU usage is way too high. Please correct this in the next update.

Photoshop 2023 and Photo AI plugin has started crashing Photoshop again.

Bug: Topaz Photo AI 1.1.0 loses GPS location metadata when storing result images.

I see many others are reporting here the same issues as me… I just got an email from them.
Their solution is to not use the update and roll back to a previous version.

Problem: Doesn’t work.
Topaz solution: Don’t use it!

Got it.

This is one of those cases of “you had one job…”, Photo AI is a very specialised tool and everything is malfunctioning! The slider is wrong, program keeps crashing, eye colour is wrong, denoise is on despite having it off, faces are warped, awful artifacts are introduced, sharpen does nothing…

Imagine having paid for Photoshop, but only half of the tools work and it will randomly work, randomly not work. That’s what this is. Amazing potential results, but this is unacceptable.

I have a M1 MBP and I can only get 1.0.7 to work on my computer. Also, RAM and storage are NOT an issue. I have tried updating to 1.0.8-1.1 and anything above 1.0.7, it will not work. I have tried deleting the plist as they suggested. I have tried allowing full disc access. I have rebooted. It is bad, it just freezes. I guess I am stuck using 1.0.7. :frowning:

Are you using windows or Mac?.

Please post your computing environment here: OS, RAM, and GPU with dedicated vRAM and driver version.

I see others have reported it as well. Wanted to confirm I also see the strange placement of face detection boxes on the preview. The top-right navigation view shows the correct position. Face recovery seems to work, so it’s just showing wrong coordinates on preview.

Support advises you to roll back as a stopgap solution so you stop experiencing crashes ASAP. I assure you we’re always working hard to fix bugs, prioritizing crashes, that get reported.

Same deal here- face box is off in left field. AI has to be difficult to debug. My processing times vary a lot. Sometimes I think the latest versions are slower, other times faster. Still looking for an official answer about what the subject mask does and doesn’t affect.

1 Like

Face box should be fixed in latest version here.

The subject mask currently informs autopilot blur/noise detection and applies to the Sharpen filter. You can toggle within the filter settings:



v1.1.1 available.