| name | generated-video-director |
| description | Specify inserted generated-video clips for the composer pipeline. Use when a transcript section needs a generated visual metaphor, illustrative clip, b-roll, product-style animation, or cinematic insert that must match the global style and neighboring shots. |
Generated Video Director
Overview
Write generated-video briefs that serve the lesson and can be assembled back into the Remotion timeline without feeling like a different video.
Use Generated Video When
- The idea is spatial, physical, emotional, or metaphorical.
- HTML motion would be too literal or too flat.
- A short insert can make the concept easier to remember.
- The shot can preserve style, color, and framing from the surrounding video.
Avoid Generated Video When
- The viewer needs exact text, UI, math, code, or diagrams.
- The presenter or Hyperframes atom can explain it more clearly.
- The generated clip would introduce a new palette, camera language, or tone.
- A deterministic R3F three-effect (glass slab, ceramic card, floating tablet) would carry the same meaning with frame-accurate composition. Prefer three-effect over generated b-roll when the goal is a labeled surface or a section handoff rather than evocative motion. See
.agents/skills/three-effects-composition.
- A real captured artifact (a UI capture, a product screen) would carry the meaning more honestly than a synthesized clip. Prefer the real artifact textured onto a surface; generated b-roll is a last resort, never AI-slop filler. When you do generate, ground the clip in a concrete reference move and hold the
/render-lab-design-system quality bar — it must match palette, camera language, and tone, not introduce a new one.
Output
Return a generated-clip brief with:
- purpose
- duration
- prompt
- negative prompt
- first frame and last frame requirements
- style constraints from global spec
- required transition handles
- whether it can run in parallel
Use references/generated-video-brief.md as the template.