Skip to main content

Sequence Planning

Nine tools for arranging scenes into timed sequences, generating and comparing variants, working with sequence archetypes, and using style packs and brief templates.


plan_sequence

Plan a sequence from analyzed scenes and a style pack. Decides shot order, hold durations, transitions, and camera overrides. Scenes must already carry metadata — run analyze_scene first. Supports per-scene style blending via metadata.style_override, and beat-synced editing when you pass beats from analyze_beats.

  • Name
    scenes
    Type
    object[]
    Description

    Required. Array of scene objects with metadata (content_type, visual_weight, motion_energy, intent_tags). Set metadata.style_override on individual scenes to blend style packs per scene.

  • Name
    style
    Type
    string
    Description

    Required. Default style pack for the sequence — prestige, energy, dramatic, minimal, intimate, corporate, kinetic, or fade. Per-scene overrides take precedence.

  • Name
    beats
    Type
    object
    Description

    Beat analysis data from analyze_beats. When supplied, scene durations snap to beat boundaries and camera intensities match audio energy.

  • Name
    duration_target_s
    Type
    number
    Description

    Target total duration in seconds (from the brief). Style-pack hold durations are scaled proportionally to land within ±0.5s, with each scene's own duration_s as the floor. If the floor sum exceeds the target, durations stay at the floor and notes.duration_target.warning is set. Omit to use raw style-pack durations.

  • Name
    preserve_source_order
    Type
    boolean
    Description

    Default true: keep the input scene order (upstream already establishes narrative order). Set false to let the planner reorder by intent buckets / energy curve — the rewrite is logged in notes.ordering_rationale and notes.ordering_mode.

Returns: A valid sequence manifest with scene order, durations, transitions, camera overrides, and editorial notes.

Try asking your AI:

"Plan a sequence from these 5 analyzed scenes with the prestige style pack"

"Plan a beat-synced sequence using these scenes, energy style, and these beats"


plan_variants

Plan multiple sequence variants from the same scenes with different styles. Each style produces an independent manifest. Use this to generate A/B choreography options for comparison.

  • Name
    scenes
    Type
    object[]
    Description

    Required. Array of analyzed scene objects with metadata (run analyze_scene first).

  • Name
    styles
    Type
    string[]
    Description

    Required. Array of style pack names to generate variants for. Minimum 2 styles.

  • Name
    sequence_id
    Type
    string
    Description

    Base sequence ID. Each variant receives a suffixed ID (e.g., seq_hero__prestige).

  • Name
    duration_target_s
    Type
    number
    Description

    Target total duration in seconds, applied to every variant (see plan_sequence).

  • Name
    preserve_source_order
    Type
    boolean
    Description

    Default true: keep input scene order across all variants. Set false to let the planner reorder (logged per variant).

Returns: An array of variants — one manifest per style — ready to feed into compare_variants.


compare_variants

Score and rank multiple sequence variants. Evaluates each variant across pacing, variety, flow, and adherence dimensions. Use after plan_variants to pick the best choreography.

  • Name
    variants
    Type
    object[]
    Description

    Required. Array of variant objects from plan_variants output — each shaped as { variant_id, style, manifest }.

  • Name
    scenes
    Type
    object[]
    Description

    Required. The same analyzed scene objects used for plan_variants.

Returns: Ranked results with per-dimension comparison (pacing, variety, flow, adherence) and an overall winner.

Try asking your AI:

"Compare these plan_variants outputs against my scene array"


evaluate_sequence

Score a planned sequence manifest against style rules and cinematography principles. Handles per-scene style blending — scenes with metadata.style_override are scored against their override pack.

  • Name
    manifest
    Type
    object
    Description

    Required. Sequence manifest from plan_sequence (must have a scenes array).

  • Name
    scenes
    Type
    object[]
    Description

    Required. The same analyzed scene objects used for plan_sequence.

  • Name
    style
    Type
    string
    Description

    Required. Default style pack to evaluate against. Per-scene metadata.style_override values take precedence.

Returns: Pacing, variety, flow, and adherence scores (0-100) with findings per dimension.

Try asking your AI:

"Evaluate this manifest against the prestige style pack"


validate_manifest

Validate a sequence manifest against camera guardrails. Checks speed limits, acceleration easing, jerk/settling, lens bounds, and personality boundaries for each scene. Returns a PASS/WARN/BLOCK verdict with per-scene diagnostics.

  • Name
    manifest
    Type
    object
    Description

    Required. Sequence manifest from plan_sequence (must have a scenes array).

  • Name
    personality
    Type
    string
    Description

    Required. Personality to validate against — cinematic-dark, editorial, neutral-light, or montage.

Returns: PASS / WARN / BLOCK verdict plus per-scene diagnostics covering speed, easing, jerk, lens bounds, and personality boundaries.

Try asking your AI:

"Validate this manifest against the editorial personality before I render"


recommend_sequence_archetype

Recommend a sequence archetype (multi-scene recipe) for a given output type. Returns scene roles, transitions, camera progression, pacing profile, and recommended primitives.

  • Name
    output_type
    Type
    string
    Description

    Required. What kind of video — brand-teaser, feature-reveal, onboarding-explainer, launch-reel, testimonial-cutdown, social-loop, or a freeform description.

  • Name
    personality
    Type
    string
    Description

    Filter archetypes by personality compatibility — cinematic-dark, editorial, neutral-light, or montage.

  • Name
    duration_s
    Type
    number
    Description

    Target duration in seconds. Helps narrow archetype selection.

Returns: The recommended archetype with scene roles, transitions, camera progression, pacing profile, and recommended primitives.

Try asking your AI:

"Recommend an archetype for a 30-second brand-teaser in cinematic-dark"


instantiate_sequence_archetype

Generate a manifest skeleton from an AI demo sequence archetype (prompt_to_answer, brief_to_board, query_to_report, upload_to_insight). Returns pre-configured scenes with timing, transitions, and camera intent.

  • Name
    archetype_slug
    Type
    string
    Description

    Required. Archetype slug (e.g., prompt_to_answer, brief_to_board).

  • Name
    personality
    Type
    string
    Description

    Personality — cinematic-dark, editorial, neutral-light, or montage.

  • Name
    duration_s
    Type
    number
    Description

    Target total duration in seconds.

  • Name
    content_hints
    Type
    object
    Description

    Optional content hints keyed by scene role.

Returns: A manifest skeleton with pre-configured scenes, timing, transitions, and camera intent.

Try asking your AI:

"Instantiate the prompt_to_answer archetype for a 25-second editorial demo"


get_style_pack

Retrieve a style pack definition. Ten packs are available: prestige, energy, dramatic, minimal, intimate, corporate, kinetic, fade, analog, and documentary.

  • Name
    name
    Type
    string
    Description

    Required. The style pack name to retrieve.

Returns: Style pack definition with cut cadence, camera preferences, transition weights, and timing biases.

Try asking your AI:

"Show me the prestige style pack definition"


list_brief_templates / get_brief_template

Browse and retrieve brief templates. Five templates are available for common video intents.

list_brief_templates takes no parameters and returns all templates with their IDs, names, descriptions, default style packs, and suggested scene counts.

    No parameters.
  • Name
    template_id
    Type
    string
    Description

    Required. Brief template ID (e.g., product-launch, brand-story, tutorial).

Returns: The full template with section structure, suggested layouts/content types per section, defaults (style pack, tone, duration), and an example brief.

Try asking your AI:

"What brief templates are available?"

"Show me the product-launch brief template"

Was this page helpful?