Firefly Text to Video needs work.

Prompt: This should appear as a still photo with almost no motion. Only allow a single blink per subject and extremely slight breathing. No head turns, no body movement, no camera movement. Preserve exact facial structure and composition. No deformation or drift.

Video: people gesticulating wildly.
 

Credit based system should be based on download of assets, not iterations due to a system that can’t follow direct orders. I’m chewing through montly credits trying to get one usable clip.