Frames from a model, or a 3D scene an agent builds for you?
Instead of generating video frames directly, some people now let an AI agent build the whole scene in a 3D tool: models, lights, camera and the render. Afterwards it is an ordinary project you can open and change.
Would you rather direct a model that paints frames, or one that builds a scene you can still fix by hand? What would you want to adjust yourself first?