1. Use a stable prompt order
Describe subject and action first, then camera, environment, lighting and sound. A repeatable order makes failed generations easier to diagnose.
CORE FORMULAEffective cost = attempt cost ÷ acceptance rate
2. One motion goal per test
Do not combine several camera moves and complex character actions in the first test. Lock one variable, observe the result and iterate.
3. Write audio as a separate block
Specify dialogue, ambience and sound effects separately. This makes it clear whether a failure came from scene direction or audio direction.
Models worth comparing next
Frequently asked questions
How long should a prompt be?
Long enough to define constraints, but short enough that each instruction can be evaluated.
Should prompts include negative instructions?
Use a short, specific exclusion list only for recurring defects.
Next step
Build your own workload plan, then save candidate models for comparison. Your plan stays in your browser.
Open workload planner →