{
  "title": "Scene Director Schema",
  "description": "Schema for describing camera, actors, and audio in a structured way for T2V/I2V prompt pipelines.",
  "type": "object",
  "additionalProperties": false,
  "required": ["camera", "actors", "audio"],
  "properties": {
    "camera": { "$ref": "#/$defs/camera" },
    "actors": {
      "type": "array",
      "minItems": 1,
      "items": { "$ref": "#/$defs/actor" }
    },
    "audio": { "$ref": "#/$defs/audio" }
  },
  "$defs": {
    "camera": {
      "type": "object",
      "additionalProperties": false,
      "required": ["composition", "movement", "position", "angle", "framing", "focus", "lens", "notes"],
      "properties": {
        "composition": { "$ref": "#/$defs/composition" },
        "movement": { "$ref": "#/$defs/movement" },
        "position": { "$ref": "#/$defs/position" },

        "angle": { "$ref": "#/$defs/angle" },
        "framing": { "$ref": "#/$defs/framing" },
        "focus": { "$ref": "#/$defs/focus" },
        "lens": { "$ref": "#/$defs/lens" },

        "notes": {
          "type": "string",
          "description": "Optional director notes for the camera."
        }
      }
    },

    "position": {
      "type": "object",
      "additionalProperties": false,
      "required": ["description"],
      "properties": {
        "description": {
          "type": "string",
          "minLength": 1,
          "description": "Natural language description of physical camera placement (e.g., 'Eye level, 2m from chef, slightly camera-left')."
        }
      }
    },

    "actor": {
      "type": "object",
      "additionalProperties": false,
      "required": ["id", "description", "action"],
      "properties": {
        "id": {
          "type": "string",
          "pattern": "^[a-z][a-z0-9_]*$",
          "description": "Stable identifier used by dialogue.actorId (snake_case recommended)."
        },
        "description": {
          "type": "string",
          "minLength": 1,
          "description": "Who/what the actor is (appearance, age range, vibe, clothing)."
        },
        "action": {
          "type": "string",
          "minLength": 1,
          "description": "What the actor is doing in the shot."
        }
      }
    },

    "audio": {
      "type": "object",
      "additionalProperties": false,
      "required": ["dialogue", "music", "sfx"],
      "properties": {
        "dialogue": {
          "type": "array",
          "items": { "$ref": "#/$defs/dialogueLine" },
          "description": "Ordered spoken lines. Empty array allowed if no dialogue."
        },
        "music": {
          "type": "string",
          "description": "Optional music direction (genre, mood, instrumentation)."
        },
        "sfx": {
          "type": "array",
          "items": { "type": "string" },
          "description": "Optional sound effects directions."
        }
      }
    },

    "dialogueLine": {
      "type": "object",
      "additionalProperties": false,
      "required": ["actorId", "line"],
      "properties": {
        "actorId": {
          "type": "string",
          "pattern": "^[a-z][a-z0-9_]*$",
          "description": "Must match an actors[].id."
        },
        "line": {
          "type": "string",
          "minLength": 1,
          "description": "Spoken line."
        }
      }
    },

    "composition": {
      "type": "string",
      "description": "Primary shot size / composition label.",
      "enum": [
        "extreme_wide_shot",
        "wide_shot",
        "full_shot",
        "medium_long_shot",
        "medium_shot",
        "medium_close_up",
        "close_up",
        "extreme_close_up",

        "two_shot",
        "three_shot",
        "group_shot",

        "over_the_shoulder",
        "point_of_view",
        "insert_shot",
        "cutaway",
        "reaction_shot",
        "establishing_shot",

        "centered",
        "rule_of_thirds_left",
        "rule_of_thirds_right",
        "symmetrical",
        "asymmetrical",
        "negative_space_heavy",

        "foreground_framing",
        "silhouette",
        "profile_shot"
      ]
    },

    "movement": {
      "type": "string",
      "description": "Primary camera movement style (single label).",
      "enum": [
        "static",

        "pan_left",
        "pan_right",
        "tilt_up",
        "tilt_down",
        "roll",

        "push_in",
        "pull_out",
        "slow_push_in",
        "slow_pull_out",
        "fast_push_in",
        "fast_pull_out",

        "zoom_in",
        "zoom_out",
        "slow_zoom_in",
        "slow_zoom_out",
        "snap_zoom_in",
        "snap_zoom_out",
        "digital_zoom",

        "dolly_in",
        "dolly_out",
        "dolly_left",
        "dolly_right",
        "tracking_forward",
        "tracking_backward",
        "tracking_left",
        "tracking_right",

        "follow_shot",
        "lead_shot",

        "crane_up",
        "crane_down",
        "jib_up",
        "jib_down",
        "boom_up",
        "boom_down",

        "handheld",
        "subtle_handheld",
        "shaky_handheld",
        "documentary_handheld",
        "steadicam",

        "orbit",
        "arc_left",
        "arc_right",
        "spiral_in",
        "spiral_out",
        "parallax_move",

        "whip_pan",
        "whip_tilt",

        "drone_static",
        "drone_push_in",
        "drone_pull_out",
        "drone_orbit",
        "drone_flyover",
        "drone_drop_down",

        "speed_ramp_in",
        "speed_ramp_out",
        "invisible_cut_push",
        "motion_blur_transition"
      ]
    },

    "angle": {
      "type": "string",
      "description": "Optional camera angle.",
      "enum": [
        "eye_level",
        "low_angle",
        "high_angle",
        "bird_eye_view",
        "worm_eye_view",
        "dutch_angle"
      ]
    },

    "framing": {
      "type": "string",
      "description": "Optional subject framing emphasis.",
      "enum": [
        "centered",
        "off_center",
        "symmetrical",
        "asymmetrical",
        "negative_space_left",
        "negative_space_right",
        "foreground_framing",
        "background_dominant"
      ]
    },

    "focus": {
      "type": "string",
      "description": "Optional focus / depth-of-field behavior.",
      "enum": [
        "deep_focus",
        "shallow_focus",
        "soft_focus",
        "macro_focus",
        "rack_focus_foreground_to_background",
        "rack_focus_background_to_foreground",
        "tilt_shift"
      ]
    },

    "lens": {
      "type": "object",
      "description": "Optional lens metadata (kept flexible; not everyone wants to be a DP).",
      "additionalProperties": false,
      "required": ["focalLengthMm", "aperture", "notes"],
      "properties": {
        "focalLengthMm": {
          "type": "number",
          "minimum": 1,
          "description": "Focal length in millimeters."
        },
        "aperture": {
          "type": "number",
          "minimum": 0.1,
          "description": "F-number (e.g., 1.2, 2.8, 5.6)."
        },
        "notes": {
          "type": "string",
          "description": "Free-form lens notes (e.g., 'anamorphic', 'vintage', 'soft edges')."
        }
      }
    }
  }
}