What feature is still missing from today's AI video generators?

If you could add one feature to an AI video platform, what would it be? Better consistency? Longer videos? Realistic avatars? Smarter editing?