Generate Script

AI-powered multi-scene script generation with cinematography details, character actions, and structured scene breakdowns.

Overview

The Generate Script node uses Gemini Flash to produce a structured, multi-scene script from a text prompt. Each scene includes a visual description, action, mood, duration hint, image prompt, and optional cinematography details (shot type, camera angle, camera movement), dialogue, location metadata, music mood, and sound effects. The output is a fully structured GeneratedScript object that can feed downstream image generation and video composition nodes.

Configuration

Field	Type	Default	Description
Provider	ScriptProvider	"gemini"	AI model provider for script generation
Model	string	"gemini-2.5-flash"	Specific model version
Scene Count	number	5	Number of scenes to generate
Style Guide	string	""	Optional style directions that influence the visual and narrative style of the script
Structure	"freeform" \| "8-step" \| "custom"	"freeform"	Script structure template. Freeform allows the AI to decide pacing; 8-step follows a classic narrative arc
Tone	string	""	Optional tone descriptor (e.g., “cinematic”, “playful”, “dark”, “documentary”)
Target Length	number	60	Target total duration in seconds for the entire script

Inputs & Outputs

• Input: in – text prompt describing the desired script topic, story, or concept
• Output: scenes – structured GeneratedScript object containing title, total duration, and array of ScriptScene objects

ScriptScene Fields

| Field | Type | Description | |——-|——|————-| | sceneNumber | number | Sequential scene index | | sceneName | string | Short name for the scene | | visualDescription | string | Detailed description of what is seen | | action | string | What happens in the scene | | mood | string \| string[] | Emotional tone(s) | | durationHint | number | Suggested duration in seconds | | imagePrompt | string | Ready-to-use prompt for image generation | | characters | ScriptSceneCharacter[] | Characters with name, description, mood, action, position | | dialogue | ScriptSceneDialogue[] | Spoken lines with speaker, text, emotion | | location | ScriptSceneLocation | Name, description, time of day, weather, lighting | | cinematography | ScriptSceneCinematography | Shot type, camera angle, camera movement | | musicMood | string | Suggested background music mood | | soundEffects | string[] | Suggested sound effects |

Best Practices

• Provide a clear, specific prompt. “30-second product ad for a fitness app showing morning routine” produces better results than “make a video.”
• Use the Tone field to set the emotional register. It is applied globally across all scenes and helps maintain consistency.
• Set Scene Count based on your target duration – roughly one scene per 5-10 seconds works well for most video formats.
• The Style Guide field is useful for maintaining visual consistency. Include details about color palette, era, or visual references.
• Use the 8-step structure for narrative-driven content (stories, ads with arcs) and freeform for informational or documentary-style content.

Common Use Cases

• Generating multi-scene storyboards for video production
• Creating structured scripts for explainer videos
• Building scene-by-scene plans for animated content
• Producing shot lists with cinematography directions
• Generating image prompts for each scene in a video project

Tips

• The output imagePrompt per scene is designed to be fed directly into a Generate Image node. Connect the script output to downstream image nodes via the scene index.
• Scene character data includes structured fields (name, mood, action, position) that can be mapped to Character and Scene nodes for advanced workflows.
• The cinematography field provides shot type, camera angle, and camera movement suggestions that inform downstream video composition.
• Scripts can be imported into Scene nodes via the mapScriptSceneToNodeData() utility function, which maps all structured fields including characters, dialogue, locations, and cinematography.
• Target Length is advisory – the AI distributes the duration across scenes but individual scene durations may vary based on content complexity.