There are some fantastic edits of films by dedicated fans which take the same source media in different aspect ratios from different releases (VHS, DVD, Blu-Ray, widescreen, open mattes, scans, etc) and combine them to present the most comprehensive picture. Unfortunately because they don’t overlap perfectly, corners can be left lacking visual content, as represented by the orange below:
This has traditionally been addressed with ambient lighting fills, which are gradients based on the available image, but this can look jarring – it is just a blurry approximation after all.
I think filling these voids with generative outpainting is a good application of generative AI. It doesn’t detract from the art due to being confined to the very corners, and it allows the art to be more fully appreciated by reducing distraction. This is in contrast to AI conversions from 4:3 to 16:9, which I feel introduce significant alterations to the framing and context surrounding the content.