Mask AI v1.3.7

Windows: Full | Online
Mac: Full | Online

Released December 15, 2020.

New features

  • Sticky save settings for PNG, JPEG, and TIF
  • Better retaining of input image format on save
  • New and cleaner help tooltips


  • Fix crash when cancelling out of Photoshop plugin mode
  • Revert brush size changes from Mask 1.3.5 (to scale brush size better on Windows)
  • Installer stability fixes, and better open program detection

Hi @brad.ito, have a look at this thread as there is an issue with redo/undo becoming deactivated when changing Mask Compute Mode that was identified in v1.3.6 …


Detect subject and detect skyline are both causing MAI to crash with this latest version. This happens both when invoked as a plugin from Studio and also when MAI used standalone. Looks like some new bugs were introduced. I can live with this until it is fixed.

Second issue I have noticed is that the MAI mask returned to Studio is causing artefacts to appear in the part image that is covered by the white portion of the MAI mask. This is noticeable and really disappointing it makes MAI unusable as a Studio plugin. These artefacts do not appear in the native Studio masks. This bug also happened before the MAI upgrade so it not something new.

I hope both are solved at the earliest, especially the second one.


Please open Mask AI, go to Help-> Graphics info and press Copy then paste the info here as you haven’t posted any details of your technical environment.

Also go to Help-> File Logging, turn them on and upload them if there are any further crashes. They can be found by opening Help-> File Logging -> Open Log folder

Mask (10.8 KB) Studio (18.0 KB)

Thank you for the prompt response. I am attaching four sets of log files in two zip attachments (remove the .crash extension).

First set in time order, loaded MAI directly and tried to autodetect subject then crashed, second set MAI invoked as plugin by Studio and successfully manually drew trimap, third set loaded Studio then created MAI mask (as per second set), then successfully saved as Studio project (i.e. only original image plus MAI mask as layers), fourth set loaded previously saved project but Studio then crtashed (there are no MAI logs for this step).

Graphics card information looks to be in log files, but it is an AMD Radeon Pro WX 2100 with 2Gb memory. I know this is the minimum but up until earlier this morning before I upgraded MAI the above scenarios worked without any crash.

The artefacts I was getting in Studio look to be due to a very high foreground recovery setting in MAI so no issue there for the moment. I still need to understand how to use multiple MAI masks in studio since I would want some adjustments to apply to all the image, different adjustments to one part, and other different adjustments to another part. Still a learning process for me.

I’ve made some further tests with MAI. Changed to preference to CPU and subject autodetect worked correctly, perhaps slightly slower than by GPU. Repeated subject autodetect successfully a few more times on the same image. Also successfully tried autodetect sky on another couple of images.

The switched preference back to GPU followed by successful subject autodetect on the original image a couple more times. Autodetect problem seems to have been solved.

However, I am still getting a crash when opening a studio project that contained MAI layers.

Firstly your GPU doesn’t meet the minimum requirements, although you may still be able to use it:

Secondly, as you didn’t open the Graphics info. and copy/paste it here you will need to check yourself if your drivers are up to date from the AMD website.

Thirdly, open Mask AI and open an image then under File-> Preferences and click Calibrate where when you next compute a mask, Mask AI will check your hardware and select the correct processing option for your PC:

AMD drivers are up to date as at last week. Yes I know the current graphics card is on the low side, however I have another Radeon Pro with 8Gb on order that should arrive in a week’s time - will see the difference and report.

I tried to update from 1.3.3 to 1.3.7 … it says, the installer will crash due to BigSur 11.0, should update to 11.1 first. The issue with this is: I already updated to 11.1. … any suggestions?

Go to the download page on the main website and use the Full Installer:

I have the exact same issue. Also tried downloading v. 1.3.7 from topaz and it would not install - same message error - need to install macOS 11.1, which I have installed and restarted several times.

I has the same problem as well. After updating to V1.3.7 using subject detect crashed the program. I did not have this problem before the update, subject detect worked fine. Is there anyway I can undo the update? I am using Windows 10 and Mask AI in a stand alone.

Select the version you want from the category of this thread … all releases for all products are in the Releases category.

Thank you. All is well again.

1 Like

New user with 1.3.7 but crashes to desktop every time using the subject detect (the second step in the tutorial) on both Mac and Windows. Tried both CPU and GPU settings on both with same result. Graphic drivers are latest on both systems.

Nevermind - found the answer in another thread. Turns out you need to choose the landscape tutorial first and let it do a one-time calibration. Works fine after that.

I am seeing the exact same issues. I’ve submitted a ticket to support the the logs to see what comes of that.

Note everybody there is an issue with the Auto-detection actions that this workaround will fix for you:

  • firstly open the preferences and click on Calibrate

  • Open an image and draw any mask so it can do the calibration, or you can open the second tutorial and it will also calibrate

Following that the detect options will work.

This is what you have said, a workaround that treats the symptom and not the cause. In all probability it will shift device from GPU to CPU.

Proper fix is what will revert the GPU to be able to be used as was the case before this upgrade.

This is to use optimum processing, the error will occur with GPU or CPU if calibration is not run. It is an issue with that not being run first.

Also calibration will choose the optimum for processing, if you want to change it to what you like after calibration … go for it, the tutorials will still run.