[D] what's the best approach for this use case?

Hi everyone,

I’m working on a project to help a friend who applies sun foils to building windows. Customers often want to see a preview of how the building will look with the foils, and right now, they’re manually photoshopping these images.

I’m new to AI and experimenting with different tools. My goal is to automate this process so we can input a building photo and get an output with darkened windows, simulating the foil effect.

Should I fine-tune a model for this specific task, or would using something like ControlNet be more effective? Any advice on the best approach would be greatly appreciated!