Shot List Generator Skill

Purpose

Convert screenplay scenes into detailed, machine-readable shot specifications that enable automated image generation.

Trigger

All reference images generated (CHARACTER_REFS, LOCATION_REFS complete).

Inputs Required

• •SCRIPTS/SCRIPT_EP{{XX}}.md
• •EP{{XX}}_SCENELIST.md
• •CANON_DB.json (with reference paths)
• •STYLEGUIDE_VISUAL.md

Outputs Produced

• •SHOT_LIST_EP{{XX}}.json - Machine-readable shot specifications

Shot Specification Schema

Each shot contains:

{
  "shot_id": "EP01_SC03_SH02",
  "scene_id": "SC03",
  "shot_number": 2,
  "shot_type": "medium",
  "duration": 4.5,

  "content": {
    "description": "Alice reviews case files while Bob paces nervously",
    "characters": ["ALICE_CHEN", "BOB_MARTINEZ"],
    "character_actions": {
      "ALICE_CHEN": "seated at desk, looking at files",
      "BOB_MARTINEZ": "standing, pacing in background"
    },
    "dialogue": "ALICE: Three connections..."
  },

  "visual": {
    "location": "PRECINCT_BULLPEN",
    "location_area": "alice_desk",
    "time_of_day": "afternoon",
    "lighting": "harsh overhead fluorescent",
    "weather": null,
    "mood": "tense"
  },

  "camera": {
    "shot_type": "medium",
    "angle": "eye_level",
    "movement": "static",
    "focus": "ALICE_CHEN",
    "depth_of_field": "shallow"
  },

  "composition": {
    "framing": "Alice screen left, Bob screen right background",
    "foreground": "case files on desk",
    "background": "evidence board visible",
    "notes": "shallow DoF on Alice, Bob slightly soft"
  },

  "references": {
    "character_refs": [
      "CHARACTER_REFS/ALICE_CHEN/refs/alice_seated_working.png",
      "CHARACTER_REFS/BOB_MARTINEZ/refs/bob_standing_concerned.png"
    ],
    "location_ref": "LOCATION_REFS/PRECINCT_BULLPEN/refs/bullpen_alice_desk.png"
  },

  "audio_cues": ["office ambience", "paper rustling", "footsteps"],

  "generation": {
    "prompt": null,
    "status": "pending",
    "attempts": 0,
    "output_path": null
  }
}

Process

Step 1: Parse Screenplay

For each scene in SCRIPT_EP{{XX}}.md:

1. •Extract scene header (location, time)
2. •Identify characters present
3. •Break down action beats
4. •Note dialogue segments

Step 2: Determine Shot Coverage

Apply standard coverage patterns:

Dialogue Scenes:

• •Establishing/master shot
• •Medium shots of each speaking character
• •Over-shoulder shots for reactions
• •Close-ups for emotional beats
• •Inserts for important details

Action Scenes:

• •Wide for geography
• •Medium for action clarity
• •Close-ups for impact moments
• •Inserts for details

Transition Scenes:

• •Establishing shot
• •Character entering/exiting

Step 3: Create Shot Breakdown

For each scene, list shots in sequence:

1. •Establishing: Location context
2. •Coverage: Standard dialogue coverage
3. •Emphasis: Key emotional/story beats
4. •Details: Insert shots for props/reactions

Step 4: Assign Shot Types

From STYLEGUIDE_VISUAL.md shot taxonomy:

Step 5: Map References

For each shot, identify required references:

Character References:

• •Select closest pose/expression match
• •Multiple characters = multiple refs

Location References:

• •Match location + area
• •Match time of day
• •Match weather if applicable

Step 6: Define Composition

For each shot, specify:

• •Character positions (screen left/right)
• •Foreground elements
• •Background elements
• •Focus point
• •Depth of field

Step 7: Estimate Durations

Based on:

• •Dialogue length (words → seconds)
• •Action complexity
• •Emotional weight
• •Pacing requirements

Typical durations:

• •Establishing: 2-4 seconds
• •Dialogue medium: 3-8 seconds
• •Close-up reaction: 1-3 seconds
• •Insert: 1-2 seconds

Step 8: Generate Shot IDs

Format: EP{{XX}}_SC{{XX}}_SH{{XX}}

• •EP: Episode number (2 digits)
• •SC: Scene number (2 digits)
• •SH: Shot number within scene (2 digits)

Step 9: Compile Shot List

Create SHOT_LIST_EP{{XX}}.json:

{
  "meta": {
    "episode": "EP01",
    "total_shots": 127,
    "total_duration": 2520,
    "generated": "2026-01-25T10:30:00Z"
  },
  "shots": [
    { ... },
    { ... }
  ]
}

Step 10: Validate

Check:

• • All scenes have coverage
• • All characters have references mapped
• • All locations have references mapped
• • Durations sum to episode length
• • Shot IDs are unique
• • No orphan references

Coverage Guidelines

Minimum Coverage Per Scene

2-Person Dialogue:

• •1 establishing/master
• •2 medium singles
• •2 over-shoulders
• •2-4 close-ups

Group Scene (3+):

• •1 establishing/master
• •Medium of each speaker
• •Select OTS for key exchanges
• •Close-ups for pivotal moments

Solo Scene:

• •1 establishing
• •1-2 mediums
• •Close-ups for internal moments
• •Inserts as needed

When to Add More Coverage

• •High emotion scenes
• •Plot-critical reveals
• •Character transformation moments
• •Action sequences

When to Use Less Coverage

• •Transitional scenes
• •Simple information transfer
• •Montage sequences

Video Clip Considerations

Shot lists feed into clip definitions. When creating shots, consider:

Start Frame Strategy

Each video clip needs a start frame. Shots marked for frame generation will be used as start frames.

Frame Prompts vs Video Prompts

Shot prompts in the shot list are for FRAME generation (still images):

• •Describe static states, not transitions
• •No temporal language
• •Single moment, single expression

Video prompts live in CLIP DEFINITIONS, not shot lists.

Shots That Need Frames

Not every shot needs a generated frame. Mark shots appropriately:

- id: 1
  priority: required  # Frame will be generated
  notes: "Start frame for Clip 1"

- id: 2
  priority: optional  # May not need frame (covered by multi-prompt)
  notes: "Generated via Clip 1 multi-prompt"

- id: 3
  priority: required
  notes: "Start frame for Clip 2 (new location)"

Notes

• •Shot list is INPUT for shot-image-generator
• •Prompts are generated by shot-image-generator, not here
• •References must exist before shot list can be complete
• •Duration estimates guide pacing, not strict requirements
• •Coverage can be adjusted during generation phase
• •Frame prompts must be static (no transitional language)
• •Clip definitions handle video sequencing, not shot lists