veo-skill_skill

Overview

This skill exposes Veo, Google’s high-quality AI video generation models (veo-3-1 and veo-3-1-fast) for AI agents. It supports 1080p HD output, reference-image guidance, multiple aspect ratios, and intelligent audio generation. The skill is focused on creating short, producible videos quickly and reliably via Monet API endpoints.

How this skill works

The skill creates asynchronous or streaming tasks against the Monet API to generate video assets. You submit a task with model, prompt, optional reference images, duration, and aspect ratio, and the API returns a task ID; you then poll or stream task progress until completion. Output includes a hosted URL for the generated video and metadata about model, status, and progress.

When to use it

• Generate short, high-quality 1080p videos with built-in audio using Veo 3.1.
• Rapid iteration or prototypes requiring ultra-fast outputs using veo-3-1-fast.
• Turn reference images into stylized motion or guided animations.
• Produce platform-specific aspect ratios (16:9, 9:16, 1:1) for social or web.
• Automate batch video generation workflows with idempotent async tasks.

Best practices

• Always supply a unique idempotency_key (UUID) to prevent duplicate tasks on retries.
• Use reference images to anchor style and maintain visual consistency across outputs.
• Poll the async task endpoint at ~5 second intervals or use the sync streaming endpoint for real-time updates.
• Keep video duration to the model defaults (8s) for optimal quality and speed; extend cautiously.
• Upload large reference files via the files endpoint to obtain a stable URL; note uploads expire after 24 hours.

Example use cases

• Create a short product promo with voice and soundtrack generated automatically.
• Produce vertical 9:16 clips for TikTok or Reels from a single brand image.
• Iterate concept animations quickly during creative reviews using veo-3-1-fast.
• Automate personalized video messages by combining templates and reference portraits.
• Archive generated outputs with task metadata for audit and replay.