Write a shot-by-shot brief for video generation

Write a shot-by-shot brief for video generation.

**Part 1 — Invariants.** Declared once, repeated in full at the top of every shot:
- Overall look: medium, film stock or render style, color grade, aspect
- Character lock: face, hair, age, wardrobe — everything that identifies them
- Location and time of day

**Part 2 — Per shot:**
- Duration in seconds. *Keep shots short. Motion and identity degrade as clips get longer, so a sequence of short shots beats one long one.*
- What moves, and only that. Name the one subject in motion; everything else holds.
- Camera: static, or one movement. Two movements in one shot produce drift.
- What must not change from the previous shot.
- Audio, if the tool takes it.

**Part 3 — Prohibitions** — one list applied to every shot: no text, no subtitles, no logos, no watermark, no extra limbs, no crowd in the background.

Then:
1. The shots most likely to break — crowds, occlusion, hands, fast motion, anything held.
2. Where to cut between generations rather than asking for one continuous clip.
3. ⚠️ Which parts are tool-specific and will need rewriting when the model updates.

Rules:
- Restate the invariants **fully** in each shot. Abbreviating is where continuity fails.
- One action per shot. A shot doing two things does neither.
- No real people's likenesses.