INTENT — Global Objective

Block Header: [INTENT]
Alias: [GOAL]
Parent Pillar: INTENT (The “Why” & Rules)
Production Phase: Phase 1: STAGING (Pre-Production Setup)
Placement: Block 1 (The Root Definition)

1. Purpose & Functional Role

[INTENT] establishes the global identity, medium, subject context, and emotional tone of the generated scene. It serves as the primary thematic vector that guides the model’s global understanding before spatial and temporal mechanics are processed.

2. Formatting & Directorial Invariants

  • Header: [INTENT] (or [GOAL]) on its own line.
  • Introductory Mandate: Must begin with "Create ..." (e.g., “Create a…”, “Create an…”).
  • Strict 1-Sentence Length Cap: Written as exactly ONE (1) single flowing sentence (hard budget: 25–50 words). Never write multi-sentence paragraphs, lore essays, or narrative synopses.
  • Prohibited Formatting & Content:
    • Do not use key-value labels (e.g., avoid Genre: Drama, Format: 35mm).
    • Do not include plot backstories, character history, or step-by-step narrative summaries.
    • Do not include shot-by-shot action, camera movements, or timeline sequences (belongs in [EXECUTION]).
    • Do not include physical guardrails or axis rules (belongs in [LOGIC]).
    • Do not include static wardrobe palettes or lens specs (belongs in [AESTHETIC]).

3. The 4 Core Semantic Vectors (Boundaries, Not Handcuffs)

Rather than enforcing a rigid fill-in-the-blank formula, the AI director has full creative freedom to construct the sentence organically, ensuring it weaves together the 4 Core Semantic Vectors in whatever phrasing flows most naturally:

Semantic Vector Directorial Scope Concrete Examples
1. Format, Medium & Tone Cinematic medium, format, era, or genre mood 1980s found-footage horror, cinematic melancholic drama, dry indie comedy teaser
2. Setting / Macro-World Primary architectural space or physical container rainy late-night laundromat, dark forest near eerie lake, flooded cathedral ruins
3. Subjects & Characters Named characters, archetypes, or personas three investigators, Samira, an exhausted artist, Robert, an elder explorer
4. Stakes & Core Action Overarching dramatic tension or central purpose confronting a nightmarish creature, secrets slowly unravel, final quiet goodbye

The 4 Vectors Explained:

  1. Format / Medium / Tone: The cinematic medium, format, or mood (e.g., 1980s found-footage mystery horror, cinematic, melancholic drama, dry, eccentric comedy ensemble teaser).
  2. Setting / Macro World: The primary architectural space or environment (e.g., at night in a dark forest near an eerie lake, in a humid, late-night laundromat, within flooded gothic cathedral ruins).
  3. Primary Subjects & Archetypes: Named characters and their core psychological traits or roles (e.g., three investigators, Samira, an exhausted artist hiding deep depression, and Felix, her gentle friend).
  4. Core Stakes & Central Action: The overarching dramatic tension, action goal, or emotional why (e.g., confronting a massive nightmarish creature, where calculated secrets slowly unravel, sharing a final moment of quiet gratitude).

Canonical Syntactic Variations (From Verified Production Scripts):

  • Variant A: Setting-First (Atmospheric / Intimate Drama): "Create a [Tone/Medium] scene set in [Setting], featuring [Characters] [Stakes/Action]." (e.g., “Create a cinematic, melancholic drama scene set in a humid, late-night laundromat, featuring Samira, an exhausted artist hiding deep depression, and Felix, her gentle, reliable long-time friend.”)
  • Variant B: Character & Conflict-First (Action / Suspense / Dark Fantasy): "Create a [Tone/Medium] sequence featuring [Characters] as they [Action/Conflict] within [Setting], evoking [Tone/Stakes]." (e.g., “Create a cinematic dark fantasy scene featuring Robert, a battle-weary elder explorer, and Rosa, a desperate young scavenger, as they face an aquatic leviathan within flooded gothic cathedral ruins, evoking a high-stakes tone of dread.”)
  • Variant C: Teaser / Ensemble / POV Sequence: "Create a [Tone/Medium] teaser featuring [Characters] showcasing [Stakes/Themes] across [Setting]." (e.g., “Create a dry, eccentric comedy ensemble teaser featuring Evan, Robert, Rosa, Sarah, and Gabriel, showcasing stylized moments across distinct environments with deadpan performances and vintage cinematic texture.”)

4. Character References & Extension Boundaries

The baseline [INTENT] operates purely on canonical character names and evocative archetypes. Specialized platform tags and continuation hooks are handled strictly via modular extensions:

1. Baseline Core (No Production Modules Active)

Use clean, descriptive character names and archetypes without platform tags: > Create a 1980s found-footage mystery horror scene set at night in a dark forest near an eerie lake, featuring three investigators confronting a massive nightmarish creature.

2. Character Reference Module (character_reference.md)

When the multimodal character mapping module is explicitly activated: - Exclusive Landing Zone: Character reference tags (@actor, @image) and voice references ((voice: @audio)) are declared exclusively inside [INTENT]: ...featuring @actor1 as SAMIRA (voice: @audio1), an exhausted artist, and @actor2 as FELIX, her gentle friend... - Optional Dedicated Block: For large ensemble casts (5+ actors) or complex unused asset lists, character bindings may be moved into a dedicated [CHARACTERS] block before [LOGIC]. - Downstream Discipline: Downstream blocks ([LOGIC], [AESTHETIC], [OPENING], [EXECUTION]) strictly use canonical character names (Samira, Felix) without repeating @actor tags.

3. Video Continuation Module (video_continuation.md)

When continuing an existing clip: - [INTENT] opens with the continuation declaration: Create a seamless cinematic continuation of @video1 featuring Mark and Robert... - Technical frame inheritance, state carryover, and terminal-frame lock rules are placed in the dedicated [CONTINUITY PROTOCOL] block.

5. System Mechanism & Downstream Relations

  • State System Role: Acts as the Global Objective Function (\(J\)), establishing the optimization target for latent space sampling across the entire scene generation.
  • Upstream Input: Direct human directorial treatment or screen idea.
  • Downstream Interface:
    • To [LOGIC]: Supplies narrative stakes and world type, informing which physical guardrails are required.
    • To [AESTHETIC]: Sets genre, lighting mood, and character archetypes for lookbook detailing.
    • To [OPENING] & [EXECUTION]: Registers canonical character identities, ensuring downstream blocks inherit clean names without repeating @actor reference tags.
Back to top