text_to_video for prompt-only generation or image_to_video with an
owner-scoped source_creation_id from the generated image library.
audio_prompt to describe the intended sound. Query
the model catalog with the exact video mode before submitting. It returns only
the selections currently available for that workflow.
