veo-build_skill

Overview

This skill lets you create and edit videos using Google's Veo 2 and Veo 3 models via the google-genai Python SDK. It supports Text-to-Video, Image-to-Video, inpainting/masking edits, and advanced controls like frame interpolation and video extension. The workflows focus on practical prompts, precise control, and repeatable pipelines for generation and editing.

How this skill works

You authenticate a Vertex AI client with the google-genai library and call Veo model endpoints to generate or edit video assets. Veo 3 is used primarily for text-to-video and image-to-video generation, while Veo 2 handles inpainting and mask-based edits. Advanced controls provide frame interpolation, extending clips, and guiding generation with reference images or detailed prompt attributes.

When to use it

• Create marketing or product demo videos from text prompts rapidly
• Animate a static illustration or photograph into motion
• Remove or replace objects inside an existing clip via masks
• Bridge two images into a smooth animated sequence (frame interpolation)
• Extend the length or context of a short video clip
• Generate consistent assets using reference images for subjects or products

Best practices

• Authenticate with Vertex AI and set project/location environment variables before running workflows
• Design prompts with camera, lighting, and style keywords for predictable results
• Use high-quality reference images and tightly defined masks for precise edits
• Test generation settings (fps, resolution, length) on short clips before scaling
• Iterate: try multiple prompts and small variations to refine visual output

Example use cases

• Text-to-Video: produce a 10–20 second product teaser from a single descriptive prompt
• Image-to-Video: animate a brand illustration into a short looping social clip
• Inpainting: remove a boom mic or passersby from a location shoot using a mask
• Frame Interpolation: create a smooth video that transitions between two keyframes
• Video Extension: expand a 6-second scene to a longer sequence while retaining style