Director

Orchestrate AI video production for any style — talking head, b-roll, product demo, podcast, or custom. This skill chains the pipeline steps. Each step has its own skill with detailed instructions.

Inputs

Ask the user for:

1. What kind of video? — style, tone, purpose (e.g. "lofi b-roll surfing video", "talking head explaining a product", "podcast clip")
2. Character — directory with character.json containing name, promptBase, referenceSheet.url, optionally voiceId
3. Scenes — scene prompts from /prompt-writer, or describe what happens and this skill will write basic prompts. For best results, use /prompt-writer first — it applies model-specific prompt structure and camera vocabulary.
4. Model — Seedance 2, Sora 2, or Kling 3.0 (see /model-provider)

Determine Format Config

Based on what the user described, determine these settings: