Skip to content

ai-video-gen

Target Audience: Marketers, content creators, and business owners producing social ads, B-roll, product demos, and narrative shorts across multiple generative-video providers without a film crew.

  • Area: Marketing
  • Category: video
  • Level: advanced
  • Input: conversation
  • Output: video-script
  • Setup: install
  • Roles: Head of Marketing, Content Manager

Generates AI videos across multiple backends — Higgsfield (Soul, Kling, Seedance, +30 models), inference.sh (Grok, Wan, OmniHuman, Fabric, foley, +40), and Google Veo…

How you ask

  • “Generate a marketing video using Veo 3”
  • “Create a text-to-video clip for our social media”
  • “Animate this product image into a video with Seedance”
  • “Generate an AI video using Wan 2.5 for our launch”

Inputs

  • request — What to produce
  • scope — company, or a client-id under {company.clients}/ — a string with one reserved value, not a closed set
  • backend — Force a specific backend instead of letting the router pick; still subject to that backend's auth preflight (veo | higgsfield | inference-sh)
  • script — The ad or scene script — beats and dialogue, not shot direction
  • locked_assets — The reusable elements already generated and approved: character sheet, product sheet, locations, props. A shot list written before assets are locked produces prompts that cannot be reproduced
  • style_prefix — One style sentence prepended to every scene prompt. Omit to have one proposed from the assets.
  • request — What was requested

Outputs

  • auth-guidance (text)
  • company-ai-video-gen (markdown)
  • mp4 (video)
  • shotlist (markdown)
  • video-prompt (markdown)