The prompts build impressive clips.

I am more interested in how to make consistent "participants" and "environment" with multiple clips so they look like one longer clip.

Will that require same dev/coding harness and spec-kit-like engineering flow?

Like create a spec on how a movie will look like, feed contexts, create tasks so that multiple clips look congruent?