A.I. masking from prompt

[WabbyTwax76]

As a real estate photographer and editor I edit and review thousands of photos a month. Something that would be a huge time saver would be the ability to type in a text box the description of something like "Windows" and then A.I. will search the image and automatically mask out everything matching that description. Masking windows indoors and outdoors for twilight imagery is something I spend alot of time on. But taking it a step further and being able to type "Cats" or "TV Screens" would also be a great help.

[CMass]

Hey @WabbyTwax76, 

This sounds like a great idea. May I ask what the current limitation is for you?

Can you share an example of what you're trying to do?  

^CM

[WabbyTwax76]

My family runs a local real estate photography company and we edit all of our own work in bulk. We edit about 1700 photos a week. In addition to using hdr batch processing we use luminance masking tools like Lumenzia when we work with individual photos like this. The more manual way is to point to point mask out each window which can be time consuming. We capture 5 brackets so we take one of the darker exposures, layer it over and then mask the window out if we'd like to see through the window more. We also create twilight looking images from daytime photos which also requires us to sometimes need to mask out windows. Currently we use this website that will convert your photo to twilight and it will detect windows with A.I. and make them look as if they're glowing. But the quality of the rest of the A.I. render is very poor so we sometimes just only use the windows since it saves us time from masking. What would be amazing is if we could mask objects by simply typing in "Windows" and press "Detect". This would be helpful with lots of other things as well.. "Trash Cans", "Cars", "Trees", "Grass". Where the object detection tool is very helpful, it does not do windows well and sometimes can not detect what we'd like. Also with complex objects like "Fences" would be nice if it were to detect the fence and nothing in between the pickets.