# Stage S4: Character & Location Reference Design

You are creating visual reference images for characters and locations in a multi-episode microdrama series.

## Story Bible

Read: `{bible_path}`

## Visual Style

Read: `{style_path}`

{user_references_section}

## Task

For each character in the story bible, create a NanoBanana prompt to generate a reference portrait.

### Character Portraits

For each character:
1. Create a detailed NanoBanana prompt that:
   - Matches the visual style defined in style.json
   - Captures the character's appearance from the story bible
   - Shows the character front-facing, clear, well-lit
   - Includes distinctive features, clothing, and accessories
   - Specifies the art style consistently across all characters

2. Write a `description.md` for the character that includes:
   - Physical appearance (for visual reference)
   - Voice description (pitch, accent, cadence — for video generation prompts)
   - Personality notes (for directing performance)

### Location References

For key locations, create NanoBanana prompts for establishing shot images that:
- Match the visual style
- Capture the atmosphere and key visual elements
- Show the location empty (no characters) as a reference backdrop

## Output

For each character, write:
- `{character_id}_prompt.txt` — NanoBanana prompt for portrait generation
- `{character_id}_description.md` — Character reference sheet

For key locations, write:
- `{location_id}_prompt.txt` — NanoBanana prompt for location reference

Also write:
- **characters_plan.json** — Summary of what will be generated:
```json
{
  "characters": [
    {"id": "...", "name": "...", "prompt_file": "..._prompt.txt", "description_file": "..._description.md"}
  ],
  "locations": [
    {"id": "...", "name": "...", "prompt_file": "..._prompt.txt"}
  ]
}
```

- **user_message.txt** — Summary of character designs and location references planned.

## Important

- All prompts must be consistent with the style defined in style.json
- Character prompts should produce images suitable as reference photos for AI video generation
- Include the style/rendering approach in EVERY prompt for consistency
- For portraits: front-facing, waist-up or full-body, neutral background, clear features
- For locations: wide shot, establishing angle, atmospheric lighting
