Gemini Video Gen

Generate videos using Google Veo models.

Overview

Gemini Video Gen uses the Google Veo API to generate short videos from text prompts, with configurable aspect ratio (16:9, 9:16, 1:1), duration (5–8 seconds), several samples per request, and an optional person-generation policy. Video generation is a long-running job: the node submits the request, waits for it to finish, downloads the finished files and stores them as binary data on the output item. You can also submit without waiting and pick the result up elsewhere.

Category: AI
Tool Name: gemini_video_gen
Version: 1

Appearance: Icon: gemini | Color: #ffffff

Node Type

Action (Binary) — handles file/binary data operations

Input / Output

Direction	Port(s)
Input	Input
Output	Output, Error

Credentials

This tool requires Google AI (Gemini) credentials. See the Credentials Guide for setup instructions.

Parameters

Parameter	Type	Required	Default	Description
Model	options	No	veo-2.0-generate-001	The video generation model to use.
				Options: veo-2.0-generate-001 (Veo 2.0)
Prompt	string	Yes	—	Text prompt describing the video to generate. Supports expressions. If empty, falls back to the input item’s prompt, text or description field.
Options	collection	No	{}	Optional generation and polling settings.
— Aspect Ratio	options	No	16:9	Video aspect ratio.
				Options: 16:9 (Landscape), 9:16 (Portrait), 1:1 (Square)
— Duration (Seconds)	number	No	5	Duration of the generated video in seconds (5-8).
— Sample Count	number	No	1	Number of video samples to generate (1-4).
— Person Generation	options	No	—	Policy for generating people in the video.
				Options: dont_allow, allow_adult, allow_all — plus the empty default, which uses the model’s own policy
— Wait For Completion	boolean	No	true	Whether to poll until the video generation is complete. If false, returns the operation name immediately.
— Poll Interval (ms)	number	No	15000	Milliseconds between poll requests when waiting for completion.
— Max Poll Duration (ms)	number	No	600000	Maximum total time to spend polling before timing out (default: 10 minutes).
— Binary Property Name	string	No	data	Property name for the primary video binary output. Additional videos use video_1, video_2, etc. Names are case-sensitive — see the upstream node’s Binary Data panel for the exact names to use.
Include Input	boolean	No	false	Whether to include the original input item fields in the output.
Max Concurrency	number	No	3	Maximum number of items to process concurrently. Keep low for video generation due to heavy resource usage.

Output Data

One output item per input item, however many videos it contains. The first video is written to the binary property named by Binary Property Name (default data); any further samples land on video_1, video_2, video_3 beside it. Files are .mp4 unless the model returns WebM. Binary data arriving from upstream is preserved.

The output item’s JSON contains only the fields below. The input item’s JSON is merged in only when Include Input is on:

Field	Description
operationName	Identifier of the generation job Veo created for this item.
model	The model that generated the video.
status	completed when the job finished, failed when it finished unsuccessfully, pending when Wait For Completion is off and the node returned without waiting.
videoCount	How many videos were actually attached to the item. Present only when Wait For Completion is on.

With Wait For Completion off, the item carries operationName and status: "pending" and no video binary — the job is still running when the node moves on, so pair this with a later step that retrieves the result.

Reference the result downstream by expression, e.g. {{ $json.status }}, and the videos through the binary property names above.

Usage Examples

• Generate a video of a forest scene using Google Veo
• Create a portrait-format video for a social media reel
• Generate multiple video variations from a single prompt
• Create an 8-second video with a detailed nature description
• Generate a square video for social media

Example Configuration

Generate one landscape clip with the defaults:

{
  "type": "gemini_video_gen",
  "parameters": {
    "model": "veo-2.0-generate-001",
    "prompt": "A golden retriever running through a field of sunflowers at sunset"
  }
}

A vertical clip for short-form video, with no people:

{
  "type": "gemini_video_gen",
  "parameters": {
    "prompt": "A satisfying close-up of fresh fruit being sliced",
    "maxConcurrency": 2,
    "options": {
      "aspectRatio": "9:16",
      "durationSeconds": 5,
      "sampleCount": 1,
      "personGeneration": "dont_allow",
      "waitForCompletion": true,
      "pollIntervalMs": 15000,
      "maxPollDurationMs": 600000,
      "binaryPropertyName": "data"
    }
  }
}

Three variants of the same prompt, with a longer polling window:

{
  "type": "gemini_video_gen",
  "parameters": {
    "prompt": "{{ $json.sceneDescription }}, photorealistic, high detail",
    "includeInput": true,
    "maxConcurrency": 1,
    "options": {
      "aspectRatio": "16:9",
      "durationSeconds": 8,
      "sampleCount": 3,
      "waitForCompletion": true,
      "pollIntervalMs": 20000,
      "maxPollDurationMs": 900000
    }
  }
}

Submit and move on without waiting:

{
  "type": "gemini_video_gen",
  "parameters": {
    "prompt": "An abstract animation of colorful geometric shapes morphing",
    "maxConcurrency": 3,
    "options": {
      "aspectRatio": "1:1",
      "durationSeconds": 6,
      "sampleCount": 1,
      "waitForCompletion": false
    }
  }
}

Error Handling

Mode	Behavior
stop	Halts workflow on first error
continue	Skips failed items, passes successful ones through
errorPort	Routes failed items to Error output port

Tips

Gemini Video Gen calls the Google Veo API to generate short videos from text prompts, supporting aspect ratios of 16:9, 9:16, or 1:1 and durations of 5 to 8 seconds. Use this tool when a workflow needs to programmatically produce motion video content from descriptive text, including support for multiple samples per request. The tool outputs binary video file data via the binary storage layer on the main channel, with failures delivered to the error output.