video-generator_skill

Overview

This skill generates short-form AI videos with synchronized native audio using Google's Veo 3.1 API. It follows an iterative workflow: storyboard the shot, craft a concise prompt with separate audio cues, generate via the API, then refine until the result matches the concept. The tool supports fast drafts and high-quality final renders with configurable duration, aspect, and resolution.

How this skill works

You provide a single-shot prompt that specifies camera work, setting, subject action, audio layers, and optional style notes. The script calls Veo 3.1 (standard or fast) using your GEMINI_API_KEY, polls for the asynchronous result, and saves MP4 assets. Iteration is encouraged: run fast drafts to refine prompts and switch to the standard model for final outputs.

When to use it

• Create social clips, reels, or TikToks that need tight motion and loop-friendly content.
• Produce podcast or newsletter headers with ambient motion and synchronized soundscapes.
• Make short product reveals or brand teasers with cinematic lighting and slow camera moves.
• Generate ambient background loops for websites or presentations.
• Rapidly prototype concepts with fast drafts before finalizing high-resolution assets.

Best practices

• Keep prompts 150–300 characters for reliable, detailed output.
• Describe one continuous shot and one clear action; avoid scene changes within an 8-second clip.
• Write audio cues in separate sentences and keep dialogue under seven words per line for clean lip-sync.
• Start simple, iterate with the fast model, then finalize with the standard model at higher resolution.
• Always include a negative prompt to exclude text, watermarks, or unwanted artifacts.

Example use cases

• Social teaser — close-up hand action in 9:16, 4–6s for Reels or Shorts.
• Podcast header — 16:9, 8s ambient studio shot with subtle room sounds.
• Product reveal — slow orbit around an object, dramatic lighting, 6–8s at 4K for hero assets.
• Newsletter visual — static window rain loop with city hum and soft music.
• Rapid prototyping — use fast model, short duration, generate 2–4 variations to pick the best.