video-cinematographylisted

# Cinematography and consistency for generated video A generation model responds to specific technical direction far better than a vague creative brief. This skill turns a storyboard beat into that direction. In Video Studio, `video-storytelling` places the beat in the narrative arc, `minimax-h3-video` and `video-provider-capabilities` define the permitted H3 request, and `fal-ai` makes the call. This skill decides the camera and visual direction inside that H3 request. ## The self-contained-prompt test Before generating, check: could someone who has never seen the intended shot picture the subject, scene, motion, and camera work from the prompt text alone? If a human reader couldn't picture it from the words, a generation model will not render it either. Vague, mood-only language ("cinematic," "epic," "moody") fails this test even when it sounds evocative -- see "Describe the visual cause, not the emotion" below. ## The five-aspect prompt formula Video generation models reliably render a described subject and scene, but routinely fail on motion, spatial framing, and camera work when those are left implicit. Structure every generation prompt to fill all five slots explicitly, not just the ones that feel obvious: ``` Subject type + key visual attributes + how to tell apart from other subjects in frame Subject motion actions in temporal order; how subjects interact with objects and each other; group action if more than one Scen