Masking is creating completely different image in REMOVE

Can someone weigh in on why the results from Masking is giving me a completely different picture after applying REMOVE. I would like to know how this is even possible? This is very important to me and can lead to a paid project.


Read all about it in the documents, a resource available from the Help menu. Especially Guidance …

I hear what you are saying… but this is terrible situation. I’m processing my picture and it’s generating and producing a ‘Convertible Saab’ in one picture and an SUV in another.
This cannot be right, I’m getting cars and vans which are not part of my photos. I’m processing my work and I get somebody else’s pictures. There has to be something wrong here because this app is giving me picture frames. plants. cars and vans… and worst of all … a picture of small boys face aged around 7 years old… and that is very threatening to say the least.

It would be helpful if someone could send me an example of what I’m writing about here.
I’ve never heard of a professional application that takes the clients pictures and renders out something that doesn’t exist in the target. Can someone please tell me where these pictures are actually coming from in the app contents? My work is for Court!

Thank you

Take a screenshot of your settings and post it here.

Don, is there a way to get rid of the green selection Box

No, but I can’t see the mask.

Not understanding the answer…Are you referring to the Green Box?

No, but I can’t see the mask:
Did I send the correct settings. Am I missing the Mask???

The green box shows what is being looked at.

If you select what you want removed you need to mask it …

And this is the result …


I am 70, and have much experience with computers and graphics… I own every app topaz has except video ai. I’m also a web designer. I use Mask AI for masking out etc.

Your answer reflects that you may be thinking I’m a newbie…

In all of the years I’ve been working with graphics … Topaz Mask is the only masking tool I’ve ever used for graphics. I create and produce videos and do much green screen work in Final Cut! I purchased Topaz Photo AI because they said they no longer support Mask AI and suggested Photo. However, I’m a bit shocked that Photo is the replacement for
Mask AI.

I’m asking for guidance because I want to be sure that I’m not missing something and that I’m not causing the app to render out images that are not mine. For example: Using your example here, if the bench was a car … a white ford Crown Victoria … and I masked it just like you did the bench, I am getting this:

when it should be:

I hope you can understand my concern…
I do know that many don’t want to waste their time with someone that may not know that they need to mask the subject … as you referred to above. I hope this clears that point. Because this is a Supreme Court matter where this picture will be evidence… I cannot just ‘throw it out there’ and is why I offered to pay for what I’m asking for…

The white car, the van and the house have been fraudulently edited and we know that for sure… what we need is to know (for sure) is how that SUV was rendered from within
Photo AI… Only the Car and the tires were masked… I may have the snapshot of the masked state but the Snagit icons are numerous and show not pic.

Can you weigh in on this subject?

Thank you,

Firstly Photo AI does not use Mask AI and is not a replacement for that. The mask you draw in the Remove enhancement is the area to be either infilled (Erase Area) or Replaced (Keep Area). The image has been tampered with but Remove has only been around for a couple of months and is still in beta.

Secondly it looks like a fake image looking at the car parked at the back in the driveway at 90 degrees to the SUV is not the same as in the original.

The process is …

If you use Keep Area you get this where it tries to infill from detail within the green box:

If you use Erase Area you get this, this is infilled from images used for training the models (Generative AI):

That is all I can see but you would need a forensic scientist to confirm what you are saying.

Thanks Don…so the SUV is generated by photo!
What doesn’t make sense to us is why Photo would apply that SUV at all… seems non sensible since I’m processing a real picture and should not ever be seeing a training image in the results! Maybe I’m missing a point in that; I’m processing my image using a software which I paid for and I’m getting an app training image instead of a rendering of my image.

BTW, it was forensic personal who created this ‘FAKE IMAGE’. and is why this is so difficult!

Thanks for weighing in… aprreciated.


If you read the guidance it says it tries to replace the area when you want to erase and, as it only knows images that were used to train the models as a source, otherwise it wouldn’t be classified as Generative AI.

You cannot “render” an image from yours using erase area, it will only render a replacement from the surrounds if you specify keep area.

One more question Please:

If I select / mask … let’s say the car front window only…and the selection box outline is red and takes up much of the screen just cracking the roof, could that be a reasonable assumption / conclusion that the cars bounding box is actually the size of the mask selection rectangle…

Meaning: when the car was placed in position it ‘was not’ cropped but the area around the car, van and 3 arches was erased in its layer. If I place that picture in Mask AI and and do auto select … it selects only the car and van and everything else is transparent…but when I choose 'Remove BG in Pixelmator… everything is gone except the car, van and the 3 arches (2) over the garage doors and roof line… Could that be why the red box is so large when I only select the car?
I’m under the impression that the green or red box is actual to the size of the entity… and if it is so then it would prove that a layer did exist – Man made.

The image shows that the height of the 'Remove Background in Pixelmator (this pic) is equal to the height in Photo AI when choosing any part of the car… an exact match in-fact.
Make sense??

Not at all, the green box is the limit of where it is going to look.

Remove in Photo AI has nothing to do with masking. You paint the area to be removed, it won’t leave a transparent background as ordinary masking will.

I advise you to read articles on Generative AI to understand. It is NOT going to remove the background as Mask AI does.

You are reading too much into it.