Runway AI Prompt Generator

Clearly differentiate the physical actions of your subject from the trajectory of the camera. If a car accelerates down a street while the camera dollies forward alongside it, describe the vehicle's forward movement and the camera's parallel tracking path in distinct phrases. Conflating subject motion with camera travel often causes generative models to slide the camera instead of animating the subject.

Ground scenes in realistic physical simulation by directing atmospheric forces. Describe wind rippling through coats or hair, rainwater billowing off tires, swirling dust flurries, rising exhaust steam, or swaying foliage. Adding secondary environmental motion prevents scenes from appearing frozen or synthetic.

Specify trajectories and pacing with concrete spatial descriptors rather than generic adjectives. Indicate whether motion travels left-to-right, pushes inward on an axis, or arcs clockwise around a central subject. Define the cadence (e.g., slow deliberate stride, sudden acceleration, or steady cinematic deceleration) to keep motion stable.

For Runway prompting, use clear, structured natural language rather than relying on long comma-separated lists of disconnected keywords. Avoid redundant hype words such as "photorealistic, 8k, masterpiece, hyperrealism". Instead, articulate specific tangible surfaces, lighting angles, optical characteristics, and authentic physical materials.

Tailor your prompt structure to your workflow. In Text-to-Video (T2V), your prompt must define all essential visual components: subject appearance, environmental geography, lighting mood, and motion. In Image-to-Video (I2V), the source image already establishes composition, subject matter, lighting, and style; focus your prompt primarily on what moves, camera behaviors, and temporal progression over time.

Current Runway models support generation windows from 2 to 10 seconds. Direct actions that are realistically achievable within the selected duration rather than overcrowding short clips with multiple conflicting plot beats. More complex sequential motion or deliberate camera choreography may benefit from longer durations, allowing actions to unfold with natural physical cadence.

Use professional cinematic terminology to guide framing and depth. Specify lens characteristics (such as 35mm anamorphic, 85mm portrait prime, or wide landscape optics), focus behavior (deep focus vs. shallow bokeh), and mechanical camera movements (dolly push, crane elevation, orbital tracking, or static tripod) to establish visual intentionality.

When iterating on generated video clips, adjust only one variable at a time between generations. If the lighting is appealing but the camera speed is too aggressive, adjust only the camera descriptor while keeping the core visual, environmental, and subject descriptions locked. This systematic approach allows you to pinpoint exactly which prompt adjustment created the desired result.