. . . is there anywhere to direct the Distraction Removal Tool to small, specific areas and not just the entire image?
According to
this concise video guide on reflection removal and people removal by Julieanne Kost of Adobe, the function to remove reflections in Lightroom always works on the entire, uncropped image, and in the current initial version is only intended to apply to reflections appearing in photos shot through glass windows that cover all or most of the field-of-view.
The removal of reflections and the removal of people are listed together under the "Distraction Removal" heading of the Lightroom
Remove tool, but they work differently. They both are based on machine learning, but when it removes reflections Lightroom doesn't use generative AI to create new image elements. It works only with the light that was actually captured by the camera, and attempts to "untangle" the subject and the reflection:
Because windows separate different kinds of scenes (like indoor and outdoor), the images being summed by the window typically have different scene content, white balances, sharpness of focus, and so on. Also, in photos that contain reflections, if two edges cross, like the edges of the turntable and the legs of the photographer, it's probably due to a superimposed reflection, because the edges of opaque objects don't cross in nature.
The AI model behind Reflection Removal identifies and untangles these two images.*
The people removal function
does employ generative AI. It selects people in the source image that the neural network identifies as distractions—i.e., not the subject of the photograph—so the Remove tool can replace them with machine-generated image elements. The people removal function seems to be just a shortcut for identifying, grouping, and erasing "distracting" people—something that could also be done with the manual Remove tool. It creates replacement content for the erased areas within the photograph that attempts to maintain consistency with their surroundings.
———
*A detailed explanation
is available here.