How to prompt Runway Gen-4.5

How to structure your prompt

Describe what appears in the frame and how it moves, in direct natural-language sentences. A reliable pattern: [camera] shot of [subject] [action] in [environment], followed by supporting details such as subject appearance, lighting, style, environmental motion, motion timing, direction and speed. Order does not change priority. Include the most critical visual and motion components and aim for clarity rather than length, because overlong prompts can create conflicting requests. For sequenced actions, add timestamp lines after the description. Choose duration and frame rate in the Runway interface.

Length and syntax

Up to 2,000 characters. Longer prompts are not better if the extra words do not change what appears on screen.

Optional timestamp lines after the natural-language description, one per action: '[00:00 through 00:02] action'. Give each action a realistic amount of time. JSON formatting is ignored by the model, so use plain sentences. No parameter flags.

Prompt skeleton

[Camera shot and movement] of [subject with key appearance] [action] in [environment]. [Lighting]. [Style]. [Environmental motion, timing, direction and speed]. [Optional timestamp lines: [00:00 through 00:0X] action].

What it’s best at

• camera choreography
• complex, sequenced motion within one prompt
• precise timing of events
• subtle atmospheric changes
• text-to-video scenes

Known limitations

• text-to-video is limited to 16:9 on this Runway variant
• exact character continuity across separate clips is not guaranteed
• timestamp timing is approximate, not frame-exact

Avoid in your prompt

• conflicting actions
• unsupported aspect ratio
• overlong prompts without visual or motion purpose
• negative phrasing
• abstract words without visual detail