veo-reference-video_skill

Overview

This skill generates videos that maintain a consistent subject appearance by using one or more reference images with Veo 3.1 on fal.ai. It is designed for creators who need a person, character, object, or pet to look the same across every frame. The skill supports configurable duration, resolution, and optional audio generation.

How this skill works

You provide a text prompt describing the scene and action plus an array of reference image URLs showing the subject from different angles. The skill calls the Veo 3.1 reference-to-video endpoint, aligns the generated frames to the visual references, and returns a hosted MP4 URL. Optional parameters control duration, resolution (720p or 1080p), and whether audio is produced.

When to use it

• Creating videos where a specific person, character, or product must remain visually consistent
• Animating avatars or branded products using supplied design or product photos
• Producing short promo clips, social media content, or demonstrations with the same subject across scenes
• Generating pet videos that preserve a pet's distinctive features
• When you can supply clear, well-lit reference images from multiple angles

Best practices

• Include multiple reference images showing different angles and expressions to improve consistency
• Use high-quality, well-lit images with the subject clearly visible against simple backgrounds
• Write prompts that describe both the overall scene and specific subject actions or expressions
• Start with default 8s duration and 720p if you experience timeouts; use 1080p for final outputs
• Ensure reference image URLs are publicly accessible to avoid 400 errors

Example use cases

• Character animation: multiple angle photos of a character and the prompt 'runs through a forest, jumping over fallen logs'
• Product showcase: product photos and the prompt 'rotates on a display stand with dramatic lighting'
• Pet clip: pet photos and the prompt 'plays fetch in a backyard, running and jumping'
• Avatar social intro: avatar design images and the prompt 'waves hello then starts dancing'