At least for my enjoyment of the Gossip Goblin videos, the world and aesthetics are pretty critical to it. I wouldn't find it at all compelling if it was just someone in their kitchen playing it themselves. I do like those videos with people playing both parts in other contexts (comedy particularly), but for other types of work, the AI generated videos are doing something that can't really be done yourself unless you become a Blender virtuoso.

Blender "virtuoso" is pushing things a bit far. I can teach someone to keyframe a rigged model in 10 minutes, and record voice acting in another 15. All-in-all, it might actually be faster to make these skits in Blender if you exclude modelling and render time.