ai-video-prompt-pack

Create, adapt, and review AI video generation prompt packs for tools such as Gemini/Veo, Sora, Runway, Pika, Luma, and similar systems. Use when Codex is asked for video generation prompts, Shorts pro

SKILL.md

GitHub

AI Video Prompt Pack

Core Workflow

  1. Identify platform or model, output length, aspect ratio, audience, campaign goal, source assets, distribution channel, and review owner.
  2. Verify current official constraints when the platform matters: duration, aspect ratio, audio support, reference images, watermarking, upload limits, rights rules, and safety policies.
  3. Reduce the idea to one clear visual job per generated clip. For short clips, prefer one transformation, one setting, one camera move, and one CTA moment.
  4. Write the prompt with concrete subject, action, scene, style, camera, motion, pacing, lighting, audio, and final-frame instructions.
  5. Add negative prompts or avoidance notes for known failure modes: unreadable text, fake logos, distorted interfaces, generic stock visuals, shaky camera, over-busy scenes, and off-brand visual cliches.
  6. Provide timing maps, overlay copy, voiceover, post-production notes, and a regeneration prompt when the first output misses the mark.
  7. Include a post-generation review checklist before publish.

Short-Form Timing Pattern

For 8-second clips, use a simple timing map:

  • 0-2 seconds: hook or problem state
  • 2-5 seconds: transformation, reveal, or product action
  • 5-7 seconds: outcome or proof state
  • 7-8 seconds: final hold for CTA or brand mark

For 9:16 vertical videos, keep important details in the middle third and avoid placing essential text near the top or bottom safe areas.

Freshness Rule

AI video tools change quickly. Before giving tool-specific claims about Gemini, Veo, Sora, Runway, Pika, Luma, YouTube Shorts, TikTok, Instagram Reels, or other platforms, verify current official docs when practical.

Read references/ai-video-prompt-checklist.md when the user needs a complete prompt pack, a platform-specific prompt, or post-generation review criteria.

Safety Rules

  • Do not claim generated video is licensable, platform-compliant, or safe for commercial use without human and rights review.
  • Do not prompt for private likenesses, living-artist style mimicry, protected characters, copyrighted logos, or voice cloning unless rights are explicit.
  • Do not rely on generated in-video text for exact brand copy or URLs; recommend adding critical text in editing unless the tool is known to handle it reliably.
  • Require review before publishing regulated, political, medical, financial, legal, employment, crisis, or reputation-sensitive content.

Deliverable Shape

For AI video prompt packs, provide:

  • Platform, format, and known constraints
  • Creative objective and audience
  • Primary prompt
  • Negative prompt or avoidance list
  • Timing map
  • Voiceover and audio direction
  • Overlay and caption copy
  • Reference asset needs
  • Post-generation review checklist
  • Regeneration prompt if needed
  • Rights, disclosure, and human-review notes

References

  • Read references/ai-video-prompt-checklist.md for platform notes, prompt fields, timing patterns, and review gates.