AI Video Scene Director
/install ai-video-scene-director
AI Video Scene Director
Use this skill to turn an AI video idea into a production-ready prompt brief without wasting generations. It does not generate media directly. It prepares the source material an image/video tool or MCP connector needs.
Workflow
-
Clarify the output.
- Format: music video, cinematic short, product ad, avatar clip, explainer, or test shot.
- Length: total seconds and target shot count.
- Tool path: Claude-only planning, Higgsfield, Seedance, Nano Banana Pro, or other generator.
- Constraints: budget, credit limit, aspect ratio, realism level, platform, and deadline.
-
Define continuity.
- Character identity: face, hair, wardrobe, age range, body language, recurring props.
- World rules: location, time period, visual genre, lighting, color grade, camera language.
- Negative constraints: what must not change between shots.
-
Build the reference plan.
- Character sheet prompt.
- Outfit or prop reference prompt.
- Scene plate prompt.
- Combined character-in-scene reference prompt.
- Use a single reference image before moving to video whenever possible.
-
Build the shot plan.
- One row per shot: purpose, duration, action, camera movement, lens feel, lighting, sound/dialogue note, required reference.
- Mark each shot as one-take or multi-shot.
- Estimate generation attempts before running.
-
Produce prompt packs.
- Image prompt for reference generation.
- Video prompt for each shot.
- Continuity reminder appended to every prompt.
- Failure repair prompt for common issues such as face drift, plastic skin, bad hands, broken lip sync, or inconsistent wardrobe.
-
Review before generation.
- Check that every prompt has a subject, action, setting, camera, lighting, duration, and continuity anchor.
- Flag expensive or vague requests before the user spends credits.
- State which assumptions are inferred.
Output Format
Start directly with the deliverable. Do not add a persona, greeting, or process preamble.
Return:
concept: one paragraph.continuity bible: concise bullets.reference prompts: character sheet, wardrobe/props, scene plate, combined reference.shot table: compact markdown table.generation order: exact sequence to run.credit risk: low, medium, or high with reason.repair prompts: 3 to 5 targeted fixes.done criteria: what a usable result must satisfy.
Guardrails
- Do not copy prompts or private workflows from a creator's paid material.
- Do not claim guaranteed views, income, virality, or ranking.
- Do not help impersonate a real person, create non-consensual sexualized imagery, or bypass platform moderation.
- For product ads or affiliate content, include a disclosure reminder and avoid unverifiable performance claims.
- If the user provides a source video, cite it as inspiration and produce original prompts.
Reference
If source validation is needed, read references/source-evidence.md. It summarizes the public video/comment evidence behind this skill and the limits of the conversion.
- Make sure OpenClaw is installed (local or Docker)
- Run the install command in chat:
/install ai-video-scene-director - After installation, invoke the skill by name or use
/ai-video-scene-director - Provide required inputs per the skill's parameter spec and get structured output
What is AI Video Scene Director?
Plan AI video scenes from a creator's concept into character sheets, scene plates, multi-shot prompt briefs, credit-aware run plans, and review checklists fo... It is an AI Agent Skill for Claude Code / OpenClaw, with 49 downloads so far.
How do I install AI Video Scene Director?
Run "/install ai-video-scene-director" in the OpenClaw or Claude Code chat to install it in one step — no extra setup required.
Is AI Video Scene Director free?
Yes, AI Video Scene Director is completely free, licensed under MIT-0. You can download, install and use it at no cost.
Which platforms does AI Video Scene Director support?
AI Video Scene Director is cross-platform and runs anywhere OpenClaw / Claude Code is available (cross-platform).
Who created AI Video Scene Director?
It is built and maintained by Zakhar Pashkin (@zack-dev-cm); the current version is v0.1.0.