veo-mcp-server 1.1.0

MCP server for 4K video generation using Google VEO 3.1 -- text-to-video, image-to-video, video extension

VEO 3.1 MCP Server

Professional-grade MCP server for Google VEO 3.1 video generation with async job pattern, API key rotation, and multi-fallback download.

What It Does

Generates AI videos using Google's VEO 3.1 model. Supports text-to-video, image-to-video, video extension, and frame interpolation. Uses an async job pattern where generation starts immediately in the background and returns a job ID for polling.

Tools (9)

Tool	Description
veo_generate_video	Generate video from a text prompt. Supports resolution (720p/1080p/4K), aspect ratio (16:9/9:16), duration (4/6/8s), negative prompts, reference images, seed control, and batch generation (1-4 videos).
veo_image_to_video	Animate a reference image with a text prompt. Supports resolution, aspect ratio, reference images, and seed control.
veo_interpolate_video	Create a smooth video transition between a first frame and a last frame image.
veo_extend_video	Extend an existing VEO-generated video by ~7 seconds. 720p only, max 148 seconds total. Source must be from a previous VEO generation (tracks provenance server-side).
veo_check_job	Check the status of an async video generation job. Call repeatedly until status is 'completed' or 'failed'.
veo_list_jobs	List all video generation jobs and their current status.
veo_api_status	Show current API key rotation status — keys configured, active key, keys remaining.
veo_pricing_info	Show VEO 3.1 pricing per second of generated video for both standard and fast models at all resolutions.
veo_show_output_stats	Display statistics about generated videos and active/recent jobs.

Configuration

Variable	Description	Default
GEMINI_API_KEY	Primary Gemini API key (required)	--
GEMINI_API_KEY_BACKUP	Backup key for automatic rotation on 429	--
VIDEO_OUTPUT_DIR	Directory where generated videos are saved	~/veo-videos

Claude Desktop Configuration

{
  "mcpServers": {
    "veo": {
      "command": "python",
      "args": ["path/to/gemini-media-mcp/servers/veo/server.py"],
      "env": {
        "GEMINI_API_KEY": "your_key_here",
        "VIDEO_OUTPUT_DIR": "./videos"
      }
    }
  }
}

Architecture

• Async Job Pattern: All generation tools return a job ID instantly. Background threads handle the long-running Google API calls, so each MCP call returns in under 5 seconds.
• API Key Rotation: On 429/RESOURCE_EXHAUSTED, automatically switches to the backup key and retries.
• 3-Layer Download Fallback: SDK download -> URI fallback -> Wait-for-ACTIVE state. Handles Google's Files API edge cases.
• Preventive Cleanup: Deletes stuck PROCESSING files before each generation to prevent cascading 500 errors.

Models

Tier	Model ID	Use Case
Standard	veo-3.1-generate-preview	Higher quality output
Fast	veo-3.1-fast-generate-preview	Quicker generation

License

MIT