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:
in– Accepts any upstream media (images, videos, audio). Multiple connections supported.
Outputs:
composition– Scene-graph plan (JSON). Must be connected to a Render Video node for final output.Best Practices
- Write descriptive prompts that reference your upstream assets by name using
@node-namesyntax. - Reorder assets in the Asset Order panel to match their intended visual priority.
- Keep duration reasonable for the number of assets – overly long compositions with few assets produce sparse timelines.
- Connect to a Render Video node downstream to produce the actual video file.
- Use 16:9 for standard video, 9:16 for vertical/mobile content, 1:1 for social media squares.
Common Use Cases
- Assembling AI-generated images and narration into a slideshow or explainer video.
- Creating documentary-style compositions from multiple media sources.
- Building social media reels from a collection of clips and images.
- Producing scene-based video from script-generated assets.
Tips
- The composition prompt is sent to Claude Sonnet, so be specific about timing, transitions, and layering.
- Asset Order affects how the AI prioritizes which media to feature. Drag the most important assets to the top.
- The output is a plan, not a rendered video. Always chain a Render Video node downstream.
- If the AI composition does not match expectations, refine the prompt with more explicit timing cues (e.g., “show image A for 5 seconds, then crossfade to video B”).