Notes on making video models behave

What we learn about prompt grammar, model quirks and the parameters that actually move the result.