Video Composer

AI-powered scene-graph video composition from natural language prompts.

Overview

The Video Composer node uses Claude Sonnet (configurable via the model selector — any of the shared LLM registry’s models) to transform a text prompt into a structured scene-graph plan. It analyzes connected upstream assets (images, videos, audio) and arranges them into a multi-track timeline composition. The resulting plan is then rendered into a final video via the Render Video node.

Reasoning-capable models additionally show an Effort selector next to the model picker (Auto by default — the vendor default, no charge change). xhigh/max bill one tier up, same rule as every other LLM-backed node — see the Generate Text node’s Reasoning effort section for the exact formula and worked examples.

Advanced mode. Gemini models offer an Advanced mode switch. Turning it on runs the model on the provider’s own API instead of through our aggregator, which is the only place Temperature, Max Tokens and the full reasoning-depth range actually take effect — those controls appear once it is on. It bills one credit tier up, and the node’s cost badge updates immediately so you can see the change before running. On a non-Gemini model the switch is visible but disabled, with the reason shown inline.

Configuration

Field Type Default Description
Composition Prompt string "" Natural language description of the desired composition. Supports @node-name references to upstream assets.
Asset Order string[] auto Drag-reorder list of connected upstream assets to control their priority in the composition.
Aspect Ratio enum "16:9" Output aspect ratio. Options: 16:9, 9:16, 1:1, 4:5.
FPS number 30 Frames per second. Options: 24, 30, 60.
Duration number (seconds) 30 Total composition duration. Range: 1–300 seconds.
Background Color hex string "#000000" Default background color for empty regions.

Inputs & Outputs

Inputs:

Outputs:

Common Use Cases

Tips