Earlier version of Beta's Generative fill worked better
Version 25.0. 0 can't extend frame better than before. Sometimes it is even totally useless when creating seamable lines and content doesn't continue from original footage.
PS. Maybe in a future we can also a version that generates grain / noise like in a original photo when deadling with hires images. At the moment it can't and AI fills has to be fixed: create noise so fills are not seen when zooming in. Luckily fills have masks so they can be used to speed up this finalizing fix.
