# FRAMEWORK_INTENT.md — The Auteur Script North Star & Architectural Intent

> *"Control is an illusion; intent is everything."*  
> — Taruma Sakti

---

## 1. Core Thesis & The North Star

**Auteur Script is first and foremost a cognitive pre-visualization scaffold for the human director, not a pixel-control protocol for the machine.**

### 1.1 Meeting the Machine Halfway
In early AI prompting, creators attempted to exert rigid, deterministic control over every pixel. But as video models evolved reasoning layers, modern generation became a collaborative process. We provide the structural blueprint, intent, and physical constraints; the model scans the structure and executes the latent rendering.

> *"A script isn't just a description—it's a set of instructions designed to trigger an experience."*

Under this paradigm, the AI acts like a **skilled film crew**: the human director establishes coordinates and boundaries in Staging and Execution, while the model's visual planning layer calculates natural physical momentum, weight, and gap-filling.

Auteur Script exists to help the human creator **mentally simulate and direct the scene before clicking generate**. 

```
[ Human Mental Cinema ] ──(Pre-Visualize)──> [ Auteur Script Scaffold ] ──(Scan Intent)──> [ Model Latent Render ]
                                                        ▲
                                                        │ (Post-Mortem Diagnostic)
                                                        └──────── Evaluate Output Divergence ────────┘
```

### 1.2 Human Cognitive Beats Over Artificial Timestamps
Modern video models can follow exact second timestamps (`00:00-00:03`), but second-accurate timestamps are unnatural for human pre-visualization:
- Humans do not instinctively imagine cinematic pacing in exact seconds.
- Humans imagine pacing through **causal state progressions and narrative beats** ($S_0 \mapsto S_1 \mapsto S_2$).
- State-transition chains (`->`) mirror natural human cognitive sequencing while giving the generative model explicit directional flow.
- Within a single execution line, the chaining operator (`->`) establishes **multi-modal synchronization and causal priority** (e.g., optical framing $\to$ actor performance $\to$ spoken line $\to$ foley contact), which the model translates into a continuous synchronized beat interval ($\Delta t$).

### 1.3 The Dual Purpose: Pre-Visualization & Diagnostic Post-Mortem
Auteur Script serves two vital functions:
1. **Pre-Generation (Scaffold):** Separating camera optics, physical anatomy, spatial blocking, and audio into isolated dimensions forces the human brain to think like a director, cinematographer, choreographer, and sound designer.
2. **Post-Generation (Diagnostic Tool):** When an output fails to match expectations (or unexpectedly exceeds them), Auteur Script serves as a root-cause debugger:
   - *Was the issue an ambiguity or conflicting instruction in our script's staging/execution?*
   - *Or was it stochastic AI variance in latent generation?*

---

## 2. Framework vs. User Implementation Rule

When developing, auditing, or discussing Auteur Script concepts, maintain strict separation between the **Meta-Framework** and **User Implementation**:

| The Framework Governs (Invariant) | The User Decides (Flexible) |
|---|---|
| **What can be controlled:** Isolated physical & optical dimensions. | **How to format:** Full canonical tags, concise shorthand, or natural language. |
| **How state transforms:** Mathematical state logic ($S_n = f(S_{n-1} \mid \text{STAGING})$). | **Tag omission:** Omitting tags or dropping syntax when hitting platform character limits. |
| **Phase separation:** Static Staging ($S_0$) vs. Kinetic Execution ($S_1 \dots S_n$). | **Workflow entry point:** Writing Execution first, Aesthetic first, or Intent first. |
| **Platform-agnostic theory:** Keeping core concepts universal. | **Production modules:** Choosing specific module tokens (`@actor`, `{dialogue}`). |

> [!IMPORTANT]
> Never design framework concepts to be rigid or prescriptive. The framework provides the architectural coordinates; the creator freely chooses how to navigate them.

---

## 3. The 6-Point Concept Litmus Test

Before introducing, altering, or resolving disputes around any block, tag, or rule in `grammar/`, evaluate it against these six non-negotiable criteria:

1. **Single-Dimension Orthogonality**: Does this concept govern strictly **one** isolated physical, optical, or narrative dimension without stepping on existing tags?
2. **2-Phase Separation**: Is the concept unambiguously classified under **Phase 1: STAGING** (static context / first-frame anchor $S_0$) or **Phase 2: EXECUTION** (kinetic temporal transition $S_1 \dots S_n$)?
3. **Cognitive Pre-Visualization Value**: Does this concept genuinely help the human director clarify and untangle their mental picture, or is it redundant decorative jargon?
4. **Model-Agnostic Purity**: Is the concept completely independent of specific AI model hacks or vendor syntax? (Vendor-specific tokens and specialized operational workflows belong strictly in `modules/`).
5. **Graceful Omission (Minimal Overhead)**: Can a creator omit this concept in a stripped-down script without breaking the coherence of the remaining scene?
6. **Core Mathematical Consistency**: Does the concept align with foundational state equations in [`grammar/conceptual_model.qmd`](grammar/conceptual_model.qmd) ($S_n = f(S_{n-1} \mid \text{STAGING})$)?

---

## 4. Concept Intent & Negative Boundary Matrix

Use this quick-reference matrix whenever you need to clarify *why* a concept exists and *what it must never do*.

### Macro-Blocks

| Block | Directorial Question | Problem / Friction It Solves | Boundary: What It NEVER Touches |
|---|---|---|---|
| **`[INTENT]`** | *"Why does this shot exist?"* | Directionless, generic AI output; lack of dramatic subtext. | **No technical camera moves or play-by-play actions.** |
| **`[LOGIC]`** | *"What physical laws cannot break?"* | **The Reasoning Bridge:** Morphing geometry, impossible physics, lost props; primes model visual planning so execution lines can remain brief and punchy. | **No aesthetic styling, narrative prose, or timeline events.** |
| **`[AESTHETIC]`** | *"What does the world look & feel like?"* | Inconsistent palette, vague lighting, unanchored wardrobe. | **No kinetic actions or chronological time progression.** |
| **`[OPENING]`** | *"What is in frame at $t = 0$ ($S_0$)?"* | Ambiguous first frame; AI guessing where subjects start. | **No state transitions ($A \to B$) or chronological progression.** |
| **`[EXECUTION]`** | *"How does the state evolve ($S_1 \dots S_n$)?"* | Temporal drift, chaotic pacing, erratic camera cuts. | **No restating static staging parameters already in $S_0$.** |

### Key Sub-State Coordinate Tags

| Tag | Isolated Dimension | Core Problem Solved | Negative Boundary |
|---|---|---|---|
| **`[CAM]`** | Camera Optics & Motion | Model conflating camera movement with actor movement. | **Never describes actor anatomy, emotion, or dialogue.** |
| **`[ACT]`** | Physical Performance | Abstract adjectives (*"looks sad"*) producing dead faces. | **Never dictates camera optics, lighting, or set layout.** |
| **`[DIAL]`** | Spoken Dialogue & Delivery | Unsynchronized dialogue or missing audio cues. | **Never describes background music or optical camera moves.** |
| **`[AUDIO]` / `[SFX]`** | Foley & Environmental Sound | Silent or generic generative audio environments. | **Never contains spoken character dialogue.** |
| **`[MUSIC]`** | Non-Diegetic Score | Muddled audio where music fights foley/dialogue. | **Never describes on-screen sound effects.** |
| **`[BLOCK]`** | 2D/3D Spatial Set Coordinates | Subject spatial drift; teleporting across cuts. | **Never dictates eyeline vector angles (handled by `[AXIS]`).** |
| **`[AXIS]`** | 180° Eyeline & Vector Action | Characters flipping left/right orientation across takes. | **Never dictates actor staging position or movement.** |
| **`[LIGHT]`** | Dynamic Lighting Shifts | Light sources flickering or drifting unnaturally. | **Never describes static lookbook tones (handled in `[AESTHETIC]`).** |
| **`[SPEED]`** | Temporal Speed Ramping | Inconsistent frame pacing (unintended slow-mo). | **Never alters scene duration or spatial coordinates.** |
| **`[PROP]`** | Interactive Object Physics | Props disappearing, morphing, or floating mid-action. | **Never describes static background set dressing.** |