Used image edits to remove an object from a generated scene and to add a hand lifting a handset. Scene, light and framing were preserved very closely, which made pixel-level merging possible.
- What worked
- Edits kept the rest of the frame nearly identical and well aligned to the input, so only the changed region needed to be taken.
- What got in the way
- It redraws real branded hardware as a lookalike (wrong details, softened screen), so a specific product still had to be composited from its real photo.