cinematic-motion-language

Cinematic Motion Language

Purpose

This skill provides the five-pillar language system for constructing video prompts with cinematic precision. Apply it to any video generation task — atmosphere shots, product reveals, character scenes, abstract motion, sacred/spiritual visuals, or any brief where imprecise language would produce unpredictable results.

The core principle: the model understands physics, geometry, sequence, and constraint — not adjectives. Replace every vague descriptor with a physical analogy, spatial coordinate, temporal sequence, or hard rule.

The Five Pillars

• 1. Camera Contract

State the camera's behavior as a hard rule before describing anything else. The model treats the camera as a character — define it or it will improvise.

Examples:

• "Static locked-off camera. Zero movement. No pan, no zoom, no dolly, no shake."
• "Slow push-in only — 10% scale change over the full duration."
• "Single handheld drift, slight organic sway, no cuts."
• Always reinforce camera rules in the negative prompt as well.

2. Motion Physics Anchor

Give every moving element a speed reference from the physical world, not an adjective. Pair physical analogies with time-anchored measurements for maximum precision.

Speed analogy examples:

• "like dust suspended in honey"
• "like embers floating in still air"
• "like smoke through a cathedral at dawn"
• "like the surface of a lake disturbed by a single drop"

Time-anchored measurements:

• "one full revolution across the entire 10-second clip"
• "roughly 6 degrees per second"
• "the pace of a clock's hour hand — imperceptibly slow"
• "travels the full arc in 8 seconds with no pause"
• Never use "slow", "fast", "gentle", "subtle" alone — always anchor to physics or time.

3. Spatial Zoning

Divide the frame into named regions and assign explicit rules to each. This prevents the model from filling empty space with invented content.

Region naming conventions:

• "Left third / center third / right third"
• "Foreground plane / midground / background"
• "Upper half / lower half"
• "Right two-thirds / left void"