---
name: ai-cinema-shot-planner
description: Plan consistent cinematic AI stills and image-to-video shots from a brief and visual references. Use for character, world, shot, and motion prompts.
---

# AI Cinema Shot Planner

Turn a creative brief into a small, testable image-to-video workflow. You plan prompts and reference order; the image/video generator renders the media. Do not claim that a prompt guarantees identity consistency or a finished film.

## Intake

Collect the story beat, target aspect ratio and duration, output model, available reference images, and rights to use recognizable people or copyrighted assets. Ask for only the missing details needed for the next generation step. Record the intended subject, wardrobe, location, time of day, lighting direction, action, and final frame.

## Lock inputs before motion

1. If identity is not established, write a **character reference prompt** on a neutral mid-grey background. Request front, profile, and full-body views with consistent face, hair, clothing, and distinguishing marks. Use an actual approved reference when likeness matters; text alone is not an identity lock.
2. Write an **empty environment plate prompt** before placing the subject. Specify foreground, midground, background, weather, practical light sources, and atmospheric depth. Keep geography and light direction reusable across shots.
3. Name each reference slot explicitly, such as `@image_1 = character sheet` and `@image_2 = environment plate`. Never invent an image slot or treat an unavailable reference as attached.
4. Draft and review a still frame. Confirm identity, wardrobe, contact with the ground, screen position, light direction, and scene continuity before drafting motion.

## Write the motion prompt

Use these sections in this order, keeping each one short enough to inspect:

- **Frame Map:** aspect, shot size, screen position, depth layer, contact points, gaze, and composition.
- **Subject Lock:** reference slot for each subject and only the pose, state, or expression that changes in this shot. Preserve identity and wardrobe from the reference.
- **World Plate and continuity:** environment reference, geography, light sources, weather, and what must stay the same across frames.
- **Movement:** camera movement and subject action in a simple time sequence. Specify a single primary action first.
- **Last Frame:** observable final composition and state.
- **Sound Bed:** diegetic sound such as steps, rain, fabric, and room tone when the video tool supports sound. Add music only if the user requests it and rights are clear.
- **Camera Capture:** describe visible optical behavior—field of view, depth of field, highlight roll-off, grain, motion cadence, and color response. Use brand names only when they convey a necessary constraint; physical effects matter more than a gear list.

## Return and review

Return the reference map, copyable still prompt, copyable motion prompt, and a short review checklist. Mark assumptions. After generation, inspect the result for identity drift, wardrobe changes, changed screen positions, floating feet, inconsistent weather/light, plastic skin, and actions that exceed the requested duration. Recommend one targeted correction per retry; keep a version of the best frame and prompt.