Prompt engineering (for video)

Prompt engineering for video is the practice of writing text prompts that consistently produce good video outputs. The basic structure: camera move + lens + lighting + subject + mood. Example: 'Slow dolly-in, 50mm lens, golden hour. A weathered detective lights a cigarette on a porch. Tense.' Different models respond to different aspects. Sora 2 listens hardest to camera grammar; Veo 3.1 rewards lighting and lens specificity; Kling 3.0 cares most about motion verbs. Good prompt engineering also means generating multiple variations rather than retrying the same prompt — small changes often produce large differences in output.