Everything I know about
directing AI cinema,
written down to teach.

The Video Movement Vault

46 camera moves, each with a real looping video example. Watch the move, copy the prompt, then go learn why it works. This is the fastest door into the Brain.

The Video Movement Vault · 46 moves, each with a real video example

AUDIT 10/10 · quick-start funnel ✓ · takeaway cards ✓ · guided path ✓ · academy games ✓ · case library ✓ · skim mode ✓ · prompt builder ✓ · continuity checklist ✓ · trust signals ✓ · synced TOC + copy buttons ✓ · 4-part engine cards ✓ · on-site mini-games ✓ · case-file proof ✓ · frame-chaining tool ✓ · on-ramp router ✓

Philosophy: Keep It Stupid Cinematic

Before any tool, any model, any prompt: a creative constitution. Every decision in this brain descends from four beliefs.

The 2026 direction

More: human, authentic, emotional, intentional. Less: artificial, overproduced, reliant on visual tricks. AI has made strange visuals cheap, which means strangeness alone is worthless. Story must always come first; the tool only amplifies the taste of the person holding it.

The Mode System

The single most important decision happens before you write a word: which mode is this film in? Mode determines which rules are active and which are deliberately suspended. Applying luxury rules to a documentary kills it; applying documentary looseness to luxury cheapens it.

How the modes map to the 2026 visual trends

The five modes, rendered by their own laws

The same grammar drives stills and motion: these five frames were generated from this chapter's rules exactly as written, one image per mode, nothing else changed. Prompt engineering for image generation IS the brain.

Prompt Physics

The core engineering idea of this entire brain: an aesthetic is a set of independent control axes, and a prompt is a weighted parameter stack, not a description. Name the axis, choose its value, and the look becomes deliberate and transferable.

The six-axis budget

Treat every prompt as a specificity budget. Light gets the most words because it drives the most result. When output looks wrong, diagnose which axis is starved. It is almost always LIGHT or ATMOSPHERE.

The universal prompt skeleton

One skeleton, filled in order. For diffusion image models lead with light and atmosphere; lead with subject only when identity must lock.

[SUBJECT / SPACE]
+ [LIGHT: direction · quality · time · colour temp · sources]
+ [ATMOSPHERE: haze · bloom · flare · air]
+ [DEPTH: foreground / midground / background · aperture]
+ [GRADE: palette · black level · saturation · contrast]
+ [CAMERA: body · lens · stop · ISO]
+ [MOTION: video only, one clear verb]
+ [TEXTURE: true texture · organic grain · soft not plastic]
+ [NEGATIVES]

The density rule

Shorter prompts render better than longer ones. Target 280 to 400 words for a single-shot video scene; never over 600 even for multi-shot sequences. Every word must do work. When in doubt, trust the reference image to carry visual information and cut the redundant description. A bloated colour paragraph with a vague light note yields a flat image every time.

Camera, Light & Grade

The technical signature: a locked stack of camera bodies, lens choices, lighting philosophy and grading rules. Fully active in PRESTIGE and DARK, partially in CHAOS, deliberately suspended in HUMAN and ANALOG.

Camera systems by use case

The colour grade standard

• Shadows: warm, lifted blacks. Never crushed. Blacks that hold detail read as film; crushed blacks read as a LUT.
• Mid-tones: strong separation, punchy. This is where "expensive" lives.
• Highlights: controlled, never blown. Sky detail survives.
• Grain: roughly 10% overlay, organic. Grain, not digital noise.
• Output: 4K to 8K, 10-bit. Ask for "8K-in-720p" texture: every frame holds infinite detail even at delivery resolution.

Lighting philosophy (PRESTIGE / DARK)

Lighting philosophy (HUMAN override)

In HUMAN mode the rule flips completely: window light or available light IS the aesthetic. Single-source natural light is intentional and correct. Do not add motivated fill or competing sources. Authenticity over control. This is the clearest example of why the mode system exists.

Motion language

• Default: smooth gimbal or drone. No handheld shake unless deliberately overridden.
• Vehicles / aircraft: locked follow, zero shake.
• Luxury and emotion: slow push-ins. Speed is cheap; slowness is expensive.
• Energy and momentum: dynamic tracking.
• Always: rule of thirds, compositionally anchored.

The DOP reference library

Referencing a cinematographer is shorthand for an entire light philosophy. Match the mood, borrow the physics.

The Lighting Rig · animated 3D diagram

Camera Angles, Framing & Perspective

A complete vocabulary for camera position (where the camera stands relative to the subject), angle (its height and tilt), framing (how much of the subject fills the frame), and perspective (how the lens renders space). AI models respond precisely to this vocabulary; a vague "nice shot of" gets you the average of the internet.

Shot size ladder (memorize the ladder, then break it with intent)

Emotion scales inversely with distance: wide shots give context and loneliness, close-ups give intimacy and pressure. A scene that never changes rung feels flat; a scene that jumps rungs without motivation feels random.

Extreme WideEXTREME WIDE SHOT: a lone man in a long coat tiny within a vast empty concrete plaza at dusk, warm sodium lamps pooling, cool sky, human scale dwarfed by architecture.⧉ COPY PROMPT

WideWIDE SHOT: the same man full body head to toe in the dusk plaza, sodium pooling warm at his feet, environment still dominant.⧉ COPY PROMPT

MediumMEDIUM SHOT: the man waist-up, warm sodium key from left, cool dusk fill, background softly out of focus.⧉ COPY PROMPT

Close-UpCLOSE-UP: the man's face shoulders-up, warm sodium key sculpting one side, cool rim on the other, plaza melted to bokeh.⧉ COPY PROMPT

Extreme CUEXTREME CLOSE-UP: only the eyes across the frame, warm sodium reflection burning in the iris, every lash and pore resolved.⧉ COPY PROMPT

Low AngleLOW ANGLE: the man towering against the dusk sky shot from near the ground, architecture converging upward, powerful and dominant.⧉ COPY PROMPT

High AngleHIGH ANGLE: the man seen from above, small and vulnerable within the pool of sodium light, long shadow stretching.⧉ COPY PROMPT

Dutch TiltDUTCH TILT: the man walking with the horizon tilted 20 degrees, unease built purely by the canted frame.⧉ COPY PROMPT

Over-ShoulderOVER-THE-SHOULDER: past his dark coat shoulder in soft foreground toward a woman facing him, her face sharp, dialogue geometry clear.⧉ COPY PROMPT

Top-DownTOP-DOWN OVERHEAD: the man directly from above at the center of concentric plaza paving circles, coat spread like a dark mark, pure graphic geometry.⧉ COPY PROMPT

The Angle Dome · animated 3D reference

The Movement Library

Every camera movement is a sentence: what moves, along which axis, at what speed, motivated by what. AI video engines execute named movements far more reliably than described feelings. This library gives the name, the copy-ready prompt line, and what the move means emotionally.

How to prompt movement (the five slots)

• Name the move: use the canonical term (dolly, truck, pedestal, pan, tilt, roll, arc, orbit, crane, zoom). One primary move per shot.
• Direction and distance: "dolly-in from 3m to 1m", "orbit 180 degrees clockwise", "pedestal up 2m".
• Speed and easing: constant, slow creep, ease-in, ease-out, speed ramp. Say it in seconds against the runtime.
• Stability register: locked, motorized smooth, gimbal, steadicam flow, handheld breath, camcorder wobble.
• Motivation: what in the scene justifies the move (a character walking, a reveal, a threat approaching). Unmotivated movement is decoration.

▲ The full Video Movement Vault · 46 moves with real video examples now opens the page, first thing after the hero.

First-Frame Builder · build a movement prompt from your own starting image

Upload a first frame, choose the camera move, add only what the image cannot say. The output keeps camera movement, image context and actor action as separated blocks, so the prompt stays reusable when you swap the frame. Your image never leaves this page.

⬆ Tap to load your first frame
(stays on your device, never uploaded)

Camera move (46 from the vault)

Actor action · what happens

Missing scene details only (optional)

Mode overlay (optional)None · frame speaks for itselfPRESTIGEHUMANANALOGCHAOSDARK

⚙ ASSEMBLE FIRST-FRAME PROMPT

⧉ COPY FULL PROMPT

The Camera & Lens Bank

Naming a camera body, lens or film stock in a prompt is shorthand for an entire rendering pipeline: sensor character, depth behaviour, colour science and era. Use gear names as style tokens: you are borrowing the look, not the machine.

The three-token stack

A complete capture line is [BODY or FORMAT] + [FOCAL LENGTH + APERTURE] + [STOCK or COLOUR SCIENCE]. Example: ARRI Alexa 65, 85mm prime at T1.8, Kodak Vision3 500T emulation. Any one token alone still steers the model; the full stack locks it.

Focal length is emotional distance

The Character Lab

Character consistency is the hardest problem in AI filmmaking and the one that separates hobbyists from production work. The answer is a strict pipeline: lock the identity first, style it second, stage it third. Never skip steps. Never combine steps.

The consistency protocol (memorize this)

Preserve identity, facial geometry, expression baseline and skin tone across all variations. Change only pose, clothing, angle and background. Camera: Sony A1, 85mm f/1.4 at f/1.6, ISO 100. Cinematic lighting, neutral editorial colour grade, true skin texture, organic film grain. Output 4K to 8K, 10-bit. Negatives: no face morph, no background bleed, no plastic skin, no fake glow.

The character pipeline (strict order)

Step 1 · Face LockFront-facing neutral studio portrait, black camisole on seamless mid-gray, even soft key, catchlights centered, expression fully neutral, identity-reference clarity.⧉ COPY PROMPT

Step 2 · OutfitSame face, same hair: full-length on mid-gray in the project wardrobe, lookbook stance, even light, wardrobe detail crisp. Identity unchanged.⧉ COPY PROMPT

Step 3 · Six-Panel SheetA 2x3 sheet of the same character: front, three-quarter left, profile, three-quarter right, back, close-up. One identity in every panel.⧉ COPY PROMPT

Step 4 · Scene PlateOnly now the scene: same identity and wardrobe crossing a rain-slicked street at dusk under warm practicals, atmosphere and depth alive.⧉ COPY PROMPT

Video Grammar

A locked structure for AI video prompts. This grammar treats the model like a crew: it needs a call sheet, not a poem. The blocks are always delivered in the same order, every prompt, no exceptions.

The 30-second architecture

30 seconds = two 15-second segments, six shots each, 0 to 2 seconds per shot. The first two seconds are the hook: one strong visual anchor that earns the rest of the runtime.

The eight mandatory elements

Every long-form video prompt must contain all eight. Missing one is the difference between a film and a screensaver.

• Hook design: the opening two-second visual anchor, named explicitly.
• Sound design: specific texture (fabric brushing stone, distant traffic hum), never "ambient sounds".
• Music progression: rise, shift, resolve. The music has an arc, not a loop.
• Dialogue or lyric cue: a spoken line, or LYRIC CUE: [the emotion the line must hit] when lyrics carry the story.
• Rule of thirds: a compositional note per key shot.
• Acting theory: Stanislavski emotional truth (what the character wants) or Chekhov physical gesture (one telling movement).
• Lighting mastery: motivated sources and a stated contrast ratio.
• Texture and motion: 8K-in-720p detail, smooth gimbal or drone motion, cinematic live-action realism only.

The block order (locked, every prompt)

Scene & Mood: [one or two sentences: what this moment IS, dramatically]

Frame Map: [where each subject sits: left/centre/right third, foreground/midground/background, what negative space remains. Multi-shot: framing per shot.]

Subject Lock (per character): [identity anchor + body orientation + pose + state + gaze + contact points + lock-down line. Trust the reference image for wardrobe; describe only what it cannot carry.]

Cross-Frame Rules: [multi-character: never swap positions, never cross centre, never change depth; distance and screen sides held. Multi-shot: what carries across the cut.]

Movement: [character motion + micro-motion + environmental motion, flowing paragraph, per-beat timestamps inline.]

Last Frame: [the exact closing composition + on-screen text suppression line.]

World Plate: [environment, era, weather, set dressing that matters.]

Sound Bed: [diegetic audio only in this grammar: specific, physical, layered.]

Capture Realism: [film-emulated, analog warmth in highlights, controlled blacks, real fabric, real skin, real haze, real grain.]

Camera Capture: [one trimmed line: body/register, lens, movement style. Never doubled.]

Why this order works

Composition before motion: the model must know where everyone is before anyone moves. Subject Lock before Movement: identity is anchored before it is stressed. Last Frame stated explicitly: AI video drifts toward its ending, so define the ending. Camera as a single closing line: the capture spec is a stamp, not a paragraph; doubling it confuses the model.

Screen Direction & the 180-Degree Rule

The audience builds an invisible map of every scene: who stands where, who looks at whom, which way the world travels. The 180-degree rule is how you protect that map. Draw an imaginary line through the eyeline of your two subjects (the axis). Keep every camera setup on ONE side of that line. Do that, and character A always looks screen-right at character B, who always looks screen-left back. Cross it carelessly, and in the next cut they appear to swap sides, gazes point the wrong way, and the audience's map shatters without them knowing why the scene feels broken.

Why this rule matters MORE in AI film

On a real set, one crew holds the geography in their heads across every setup. In AI film, every generation is a brand-new crew that never saw the previous shot. The model has no memory of where your characters stood ten seconds ago. You are the continuity department, and the 180-degree rule is your first tool: it turns spatial logic into written law that every prompt must obey.

The four legal ways to cross the line

• Cross on camera: move the camera across the axis in one visible, unbroken move (a tracking arc). The audience travels with you, so the map updates instead of breaking.
• Cut to neutral: a shot ON the axis itself (dead frontal, or directly behind a character). Neutral shots reset the line; the next cut may land on either side.
• Let the subjects re-block: if a character physically walks to the other side within a shot, the axis re-draws itself in view, legally.
• Cutaway and return: an insert (hands, a clock, a detail) briefly releases the geography; return on the new side reads as a soft reset. The weakest of the four, use sparingly.

Enforcing the rule inside prompts

In this brain's video grammar, the rule lives in two blocks: Frame Map declares each character's screen side and gaze direction; Cross-Frame Rules makes those declarations law across shots. Copy the enforcement lines below into any multi-shot or multi-character prompt.

Cross-Frame Rules: Character A holds the LEFT third in every shot, body angled right, gaze locked screen-right toward Character B. Character B holds the RIGHT third in every shot, body angled left, gaze locked screen-left toward Character A. Camera remains on the established side of their eyeline axis for all coverage. Characters never swap screen sides, never cross the centre line, never change relative depth. Screen direction of all movement is held left-to-right across every cut. Any axis change must happen inside a single visible camera move, never between cuts.

Common breaks and their fixes

The AxisOverhead: two actors face each other, a glowing line runs between and past them. All cameras live on one side of that line.⧉ COPY PROMPT

Shot AOver-the-shoulder past his left shoulder: she looks frame-LEFT toward him. Note her gaze direction.⧉ COPY PROMPT

Shot B · ReverseReverse past her right shoulder: he looks frame-RIGHT toward her. Opposed gazes make the cut legal.⧉ COPY PROMPT

The Living Axis · animated diagram

Polished Filmmaking: Continuity & Coherence

Anyone can generate a beautiful clip. A film is different: it is clips that believe in each other. Same people, same place, same light, same emotional temperature, flowing as one continuous world. This chapter is the system for turning disconnected AI generations into a coherent film with continuity, pacing and real cinematic storytelling. It runs on five engines: asset saving, frame chaining, emotion control, light-and-lens continuity, and edit-room technique.

Engine 1 · Asset saving (the project vault)

Continuity begins before the first shot: never describe the same thing twice from words alone. Every project gets a vault of locked, reusable reference assets, and every generation pulls from that vault:

• Character assets: the face lock, the base outfit reference, and the six-panel sheet from the Character Lab. Attached to every shot the character appears in, in the same numbered order every time.
• Location plates: generate each location once as a clean master plate (wide, empty, correct time of day), approve it, save it. Every scene in that location references the plate, so the room never redecorates itself between shots.
• Prop and wardrobe references: the hero object, the vehicle, the signature jacket, each locked as its own image. Props that matter get the same treatment as faces.
• The naming discipline: assets named by project, subject and version, references listed in a fixed numbered order and tagged inline. A vault you cannot navigate is a vault you will not use under deadline.

Engine 2 · Frame chaining (video references)

The single strongest continuity technique in AI video: the last frame of the approved shot becomes the reference (or the literal first frame) of the next shot. Light, wardrobe state, geography and grade travel across the cut automatically, because the new generation starts from where the old one ended. Chain shots this way and a sequence stops being a slideshow.

1. Generate Shot A. Judge it. Approve it.
2. Export the exact final frame of Shot A as an image.
3. Shot B prompt: attach that frame as the primary reference (or as the first-frame input where the platform supports it), alongside the character assets from the vault.
4. In Shot B's Subject Lock, describe only what CHANGES from the chained frame (a new gesture, a door now open, breath fogging). Everything else is carried by the frame.
5. Repeat down the sequence. If a shot is rejected, re-chain from the last APPROVED frame, never from a rejected one.

Engine 3 · Emotion control across cuts

Faces must be continuous too. An actor cannot teleport from calm to sobbing between two cuts, and neither can your character. Build a simple emotion timeline for every scene using the Expression Engine: name the muscle register per shot, and move it one step per cut.

Scene emotion arc: [name the journey, e.g. composed → doubt → held grief → release]
Shot 1: composed baseline, still face, low blink rate.
Shot 2: first crack, inner brows lifting slightly, a swallow.
Shot 3: held grief, chin pressure upward, lips tightening, tear film forming.
Shot 4: release, one tear falling, breath breaking, gaze dropping.
Rule: each shot inherits the previous shot's register and moves it ONE step. No emotional teleporting between cuts.

Engine 4 · Light and lens continuity

• One light logic per scene: the same source direction, colour temperature and time of day stated identically in every shot's prompt. Copy-paste the scene's light line; do not rewrite it from memory.
• One lens register per scene: if coverage lives on 50mm and 85mm, it stays there. A random 24mm shot in an 85mm scene reads as a different film.
• Grade once, over everything: generate slightly flat, then apply a single grade across all clips in the edit. Per-clip grading is where coherence dies.
• Weather and state lock: if it rained in shot 3, shoulders are still damp in shot 4. State carries forward until you change it on purpose.

Engine 5 · The edit room (where clips become a film)

The pre-generation continuity checklist

□ Character assets attached, same numbered order as every previous shot
□ Location plate attached (never re-described from words)
□ Chained frame from the last APPROVED shot attached
□ Scene light line copy-pasted identically (source, temperature, time)
□ Lens register matches the scene (no focal-length strangers)
□ Screen sides and gaze directions restated (Cross-Frame Rules)
□ Emotion register = previous shot's register moved ONE step
□ State carried forward (weather, damage, wardrobe, props)
□ Last Frame written, with the next shot's match in mind
□ Sound bed continuous with the previous shot

Motion Graphics Prompting

Motion design is a different animal from live-action: here the subject is shape, type and rhythm, not light physics. The engine still wants a production document, but the blocks change: style system, element, motion verb with timing, palette discipline, background, loop behaviour.

The motion graphics skeleton

[STYLE SYSTEM: flat 2D vector / isometric 3D / paper cutout / liquid / neon wireframe / kinetic type ...]
+ [ELEMENT: the logo, word, icon, chart, character or shape]
+ [MOTION VERB + TIMING: assembles, morphs, unfolds, orbits ... over Xs, with named easing]
+ [PALETTE: 2 to 3 colours maximum, named]
+ [BACKGROUND: flat, gradient, grain, transparent]
+ [CAMERA: usually locked; state any push or parallax explicitly]
+ [LOOP: seamless loop yes/no; if yes, first frame equals last frame]

The ten commandments of motion design prompts

• Two or three colours, never more. Palette discipline is 80% of "premium" in motion graphics.
• Name the easing. Ease-in-out for elegance, overshoot-and-settle for playfulness, linear only for mechanical or data feels.
• Timing in seconds. "The word assembles over 1.2s, holds 2s, exits in 0.6s." Rhythm is the design.
• Anticipation and follow-through. A tiny counter-move before the main move, a settle after it. This is what separates animation from movement.
• One hero element per composition. Everything else is supporting geometry.
• Seamless loops must be declared. Say "seamless loop, first and last frame identical" or the engine will not close the cycle.
• Flat means flat. If you want 2D vector, negate depth: "no 3D, no shadows, no perspective".
• Transitions are named: morph, mask reveal, wipe, match cut, liquid melt, particle dissolve, page turn, shutter, iris.
• Type is a character. Give it weight, case, tracking and behaviour, not just a font vibe.
• Negative space is part of the layout. State where the emptiness lives.

Light RevealA dark emblem emerging as a sweep of golden light traces its silhouette edge, glow particles trailing the sweep, premium energy on pure black.⧉ COPY PROMPT

Particle FormationThousands of golden particles mid-swarm converging toward a crystallizing form, some still drifting free, depth through the cloud.⧉ COPY PROMPT

Liquid GoldMolten metallic gold mid-splash frozen in an elegant crown arc, viscous ribbons suspended, mirror-black surface reflecting.⧉ COPY PROMPT

Wireframe BuildAn amber holographic wireframe mid-construction, half resolved into glowing surfaces while lines and nodes still assemble the rest.⧉ COPY PROMPT

Ink BloomLuminous amber ink blooming underwater in slow organic tendrils against deep black, one cool teal wisp intertwining.⧉ COPY PROMPT

Neon Line TraceA single continuous neon line drawing itself through darkness, bright head glowing hot with a fading tail, reflections below.⧉ COPY PROMPT

The Expression Engine

The most powerful acting tool in AI film is a scientific one: the Facial Action Coding System, which breaks every human expression into numbered muscle movements called Action Units (AUs). Instead of prompting "she looks sad" (an adjective, an output), you prompt the muscles (the cause): inner brows pulled up and together, lip corners drawn down, chin trembling upward. Models render muscles far more faithfully than moods.

The core Action Units (the working set)

Emotion recipes (AU combinations)

JoyPortrait, same character: Indian woman early thirties, dark hair tied back, olive shirt, charcoal studio, soft key left, cool rim right, true skin pores, film grain. Expression JOY: genuine Duchenne smile, cheeks raised, crow's feet crinkled, eyes bright and narrowed with warmth.⧉ COPY PROMPT

GriefPortrait, same character: Indian woman early thirties, dark hair tied back, olive shirt, charcoal studio, soft key left, cool rim right, true skin pores, film grain. Expression GRIEF: inner brows raised and drawn together, lip corners pulled down, chin trembling, eyes glassing with an unfallen tear.⧉ COPY PROMPT

FearPortrait, same character: Indian woman early thirties, dark hair tied back, olive shirt, charcoal studio, soft key left, cool rim right, true skin pores, film grain. Expression FEAR: eyes wide with upper whites showing, brows raised flat and tense, lips stretched horizontally, nostrils flared.⧉ COPY PROMPT

AngerPortrait, same character: Indian woman early thirties, dark hair tied back, olive shirt, charcoal studio, soft key left, cool rim right, true skin pores, film grain. Expression ANGER: brows lowered and drawn hard together, upper eyelids raised in a glare, lips pressed thin, jaw clenched.⧉ COPY PROMPT

SurprisePortrait, same character: Indian woman early thirties, dark hair tied back, olive shirt, charcoal studio, soft key left, cool rim right, true skin pores, film grain. Expression SURPRISE: brows arched high, eyes round and open, jaw dropped softly with lips parted in an O.⧉ COPY PROMPT

DisgustPortrait, same character: Indian woman early thirties, dark hair tied back, olive shirt, charcoal studio, soft key left, cool rim right, true skin pores, film grain. Expression DISGUST: nose wrinkled, upper lip raised on one side, eyes narrowed, slight head recoil.⧉ COPY PROMPT

ContemptPortrait, same character: Indian woman early thirties, dark hair tied back, olive shirt, charcoal studio, soft key left, cool rim right, true skin pores, film grain. Expression CONTEMPT: one lip corner tightened and raised asymmetrically, one brow arched, chin lifted, eyes cool.⧉ COPY PROMPT

DeterminationPortrait, same character: Indian woman early thirties, dark hair tied back, olive shirt, charcoal studio, soft key left, cool rim right, true skin pores, film grain. Expression DETERMINATION: brows set low and steady, eyes locked on target, lips pressed with resolve, jaw forward.⧉ COPY PROMPT

TendernessPortrait, same character: Indian woman early thirties, dark hair tied back, olive shirt, charcoal studio, soft key left, cool rim right, true skin pores, film grain. Expression TENDERNESS: soft warm gaze slightly lowered, gentle closed-lip smile, relaxed brows, head tilted a few degrees.⧉ COPY PROMPT

ExhaustionPortrait, same character: Indian woman early thirties, dark hair tied back, olive shirt, charcoal studio, soft key left, cool rim right, true skin pores, film grain. Expression EXHAUSTION: heavy half-lowered eyelids, dark under-eye shadow, slack jaw, head sagging slightly forward.⧉ COPY PROMPT

HopePortrait, same character: Indian woman early thirties, dark hair tied back, olive shirt, charcoal studio, soft key left, cool rim right, true skin pores, film grain. Expression HOPE: chin lifting, eyes rising toward light above the lens, brows gently raised at center, lips parting into the first breath of a smile.⧉ COPY PROMPT

SuspicionPortrait, same character: Indian woman early thirties, dark hair tied back, olive shirt, charcoal studio, soft key left, cool rim right, true skin pores, film grain. Expression SUSPICION: eyes narrowed and cut sideways off-lens, one brow raised, lips pursed to one corner, head turned while the gaze stays locked.⧉ COPY PROMPT

Skin & Texture Realism

Realism is not a resolution number: it is evidence of life. Pores, vellus hair, uneven tone, fabric memory, dust in the air. AI defaults to airbrushed perfection because perfection is the average; you must prompt the imperfections back in, one specific token at a time.

The realism hierarchy

• Level 1, the surface: visible pores, true skin texture, no plastic, no fake glow. The minimum for any face.
• Level 2, the light response: subsurface scattering (light glowing softly through ears, nostrils and fingertips), specular highlight breakup (highlights broken by pores, never one smooth sheen), backlit vellus hair (peach fuzz rimmed by light).
• Level 3, the asymmetry: uneven skin tone, one brow slightly higher, faint redness around nostrils and eyelids, natural imperfect teeth. Symmetry is the biggest AI tell.
• Level 4, the history: freckle clusters, sun lines, a faded scar, chapped lips, knuckle creases, calluses. Skin that has lived.
• Level 5, the moment: sweat beads on the hairline, a tear film on the lower lid, wind-lifted flyaways, dust on the shoulders. The state of right now.

Pores & OilExtreme close of a cheek and nose bridge in soft window light, every pore resolved, natural oil sheen on the highlight plane, honest micro-imperfections, shallow depth, film grain.⧉ COPY PROMPT

Age & VeinsElderly hands clasped in warm lamplight, deep wrinkles, age spots and raised veins carved by raking side light, papery texture, dark falloff, film grain.⧉ COPY PROMPT

Vellus BacklightProfile edge of a young face against low golden backlight, fine vellus peach-fuzz hairs glowing individually along the jaw, subsurface warmth, film grain.⧉ COPY PROMPT

Sweat BeadsAthlete's temple and brow after effort, beads of sweat catching a hard cool key, each droplet sharp with tiny highlights, flushed skin, film grain.⧉ COPY PROMPT

Sun FrecklesFreckled shoulder and collarbone in direct summer sunlight, natural freckles sharp across sun-warmed skin, soft sky bokeh, film grain.⧉ COPY PROMPT

Rain-WetRain-wet face fragment at night, droplets and thin rivulets running down a cheekbone lit by cool neon rim with warm street fill, wet lashes, film grain.⧉ COPY PROMPT

Signature Looks

Three fully engineered aesthetic profiles. Each is decoded the same way: the DNA (remove one pillar and the look collapses), the overrides against the base signature, and the vocabulary. Profiles are closed systems: never blend two looks in one frame.

The Architecture Standard

Property and architecture film is where flat renders go to die. The standard: deep tonal range, layered light sources, real shadow depth. No flat renders, ever.

Camera: Sony A7R V, 50mm f/1.8, ISO 200. Grade: warm shadows, lifted blacks, strong mid-tone separation, film grain 10%, 8K detail. Deep tonal range, layered light sources, real shadow depth, no flat renders. Deliver as a single prompt block.

Delivery rule: architecture prompts ship as a single clean block. No tables, no breakdowns, no headers inside the prompt itself, no preamble, no explanation after. State the active mode and any override tags above the block, then just the prompt.

Pairings: KISC + Timeless Score for prestige reveals; Dynamic Timelapse + Timeless Score for masterplan and city-scale stories; drone wide at golden hour for aerial reveals.

The Sound Pairing Matrix

Music is not decoration; it is half the direction. Eight audio strategies, and the canonical picture-music pairings that work as defaults.

Canonical pairings

The Platform Map

Same idea, different grammar per engine. Compile the spec to the platform; never paste a prompt built for one engine into another.

Prompt architecture by content type

The 2026 Engine Playbook

Field intelligence from the current frontier engines, distilled into rules the Brain can execute. The Platform Map tells you the grammar family; this playbook tells you the moves that win on today's named engines. Every card follows the same anatomy: Rule, Controls what, Vocabulary, Failure mode.

Structure moves

Realism and effects moves

Workflow moves

Named-engine routing · 2026

Train the moves

Case Files

Real diagnoses from production. Each case names the symptom, the starved axis, the one lever that was changed, and the mode it ran under. This is the Debug Room's discipline applied to finished work.

DELIVERABLES THIS SYSTEM PRODUCES · brand films · architecture reels · character campaigns · event content · teaching cinema — built with it daily at FAIM for GCC clients.

The Co-Writer Method

Prompt engineering is a science on the language side too, not only the visual side. This chapter maps the beginner-to-advanced ladder of instruction craft onto directing: how to brief an AI assistant so it writes production-grade prompts with you, and how to write instructions any engine obeys. Prompt Physics governs what goes IN a prompt; the Co-Writer Method governs HOW instructions are engineered.

The instruction ladder · beginner to advanced

The Prompt Builder

The whole system as a machine. Choose the parameters; the builder assembles an engineered starting prompt from the grammar in this brain. Then do what a director does: refine one axis at a time.

Choose parameters and press Assemble.

The Sample Library

Fully written, production-shaped example prompts, anonymized and original, one per major pattern in this brain. Copy, then swap the subject for yours. Filter by category.

Scene & Mood: A night ferryman waits at an empty jetty as the last light dies over the water. The moment is patience turning into quiet resolve.

Frame Map: Ferryman on the right third, midground, facing left across open water; the jetty lamp burns in the upper-left third; the left half of the frame is negative space of dark water and sky.

Subject Lock, @image1: The ferryman from the reference. Seated on a bollard, forearms on knees, weight settled, gaze fixed left across the water, right hand loosely holding a coiled rope, boots planted on wet planks. Hold identity, face geometry and skin tone exactly; wardrobe carried by the reference; add only a damp sheen on the shoulders from sea spray.

Movement: 0 to 3s, stillness; only the lamp flame flickers and moths orbit it. 3 to 7s, he draws one slow breath, shoulders rising, and turns his head a few degrees further left. 7 to 12s, a distant boat light appears in the far-left background and he rises halfway to standing, rope tightening in his hand. Water laps continuously; a slow 10% push-in across the full runtime.

Last Frame: Ferryman half-risen on the right third, silhouette rimmed by the jetty lamp, the distant boat light dead on the left-third intersection. No on-screen text of any kind.

World Plate: A weathered wooden jetty at night, wet planks, coiled ropes, a single warm sodium lamp, calm black water, low haze on the horizon.

Sound Bed: Water lapping against pilings, the lamp's faint electrical buzz, rope creak, one distant low boat horn at 8s.

Capture Realism: Film-emulated, analog warmth in the lamp highlights, blacks lifted and holding detail, real fabric weave, real skin texture, gentle haze, organic grain.

Camera Capture: Cinema register, 55mm anamorphic character, slow motorized push-in, no shake.

MODE: PRESTIGE · TREND: Keep It Stupid Cinematic · AUDIO: Timeless Score
HOOK (0 to 2s): Extreme close-up of a craftsman's hands pressing a seal into hot wax, steam curling through a shaft of window light.

SEGMENT A (0 to 15s, six shots): 1) the wax seal hook, rule of thirds on the hands; 2) wide reveal of a stone workshop at dawn, warm practicals against cool window daylight; 3) the craftsman's face, Stanislavski intent: he wants this one to be perfect; 4) macro texture of tooled leather grain, 8K-in-720p detail; 5) slow gimbal track along the workbench, dust motes in the light shafts; 6) he lifts the finished piece toward the window, a turn: is it good enough?

SEGMENT B (15 to 30s, six shots): 7) daylight exterior, he steps into the street, score lifts; 8) emotional peak: an apprentice's face lighting up on receiving the piece, Chekhov gesture: she holds it to her chest; 9) wide payoff of the old quarter at golden hour, drone push; 10) human close-up, the craftsman's quiet half-smile; 11) resolving slow push-in on the workshop door closing; 12) final frame: the seal impression in wax filling the right third, negative space left for the mark.

MUSIC PROGRESSION: solo cello under Segment A, strings enter at 15s, resolve to a single held note at 28s.
DIALOGUE (at 14s, off-screen, low): "Again. Until it deserves the name."
SOUND DESIGN: wax hiss, leather creak, distant church bell, morning birds.
LIGHTING: two competing sources in every interior, warm practicals vs cool daylight, contrast ratio held; golden hour exteriors.
GRADE: warm lifted shadows, punchy mids, controlled highlights, 10% organic grain, 8K.
MOTION: smooth gimbal and drone only, slow push-ins for emotion.
NEGATIVES: no text overlays, no plastic skin, no fake glow, no anime, no handheld shake, no teal-orange extremes.

Photorealistic 3:4 portrait, chest-up, of a woman in her early thirties by build: oval face, high cheekbones, almond deep-brown eyes, straight full brows, softly rounded nose, natural full lips, warm medium-brown skin with a matte natural finish, a small beauty mark below the left eye. Hair: near-black with warm undertones, collarbone length, loose natural wave, centre part. Minimal makeup register, calm neutral expression with quiet confidence. Wardrobe locked to a plain black camisole, no jewellery, no styling. Mid-gray seamless studio backdrop, soft directional light from camera-left, gentle falloff, true pores and skin texture, strand-level hair detail, subtle subsurface scattering, fine theatrical grain. No environment, no props, no fashion styling, no plastic skin, no fake glow, no face stylization. Identity reference image only.

One single 16:9 frame containing a 3×2 grid of the same locked character from the attached reference, identical identity, outfit and styling in all six panels, mid-gray seamless studio, soft even directional light. Panel 1: full front body, standing relaxed. Panel 2: full back body. Panel 3: left side-profile close headshot. Panel 4: right side-profile close headshot. Panel 5: front face close headshot, neutral expression. Panel 6: detail shot of the hands and signature ring. Preserve facial geometry, skin tone, hair and proportions exactly across all panels; change only the angle. True skin texture, fabric weave visible, organic grain. No text labels, no borders drawn as UI, no face morph between panels.

A modern coastal villa at blue hour, seen from the garden's far corner across a still infinity pool. Warm interior lighting glows through floor-to-ceiling glazing, pooling gold on the terrace stone, against a deep cobalt sky holding its last light. Rich darks in the landscaping, glowing light pools along the path, cinematic contrast; the pool surface mirrors the lit facade. Foreground: soft out-of-focus palm fronds framing the lower-left; midground: the villa on the right two-thirds; background: dark sea horizon. Camera: Sony A7R V, 50mm f/1.8, ISO 200, tripod. Grade: warm shadows, lifted blacks, strong mid-tone separation, film grain 10%, 8K detail, deep tonal range, layered light sources, real shadow depth, no flat render. No HDR halos, no overexposed sky, no CGI-obvious surfaces, no people.

A slow gimbal glide through a double-height living space in late afternoon. Two competing light sources hold the frame: warm recessed ceiling practicals against cool daylight flooding from the glazing, dust motes drifting in the shafts. The camera moves at walking pace from the entry axis toward the window wall, 0 to 6s, then eases into a gentle 20-degree pan across the open kitchen, 6 to 10s. Marble reflections shift as the camera passes; sheer curtains breathe once in a draft. Sony A7R V register, 35mm, smooth stabilized motion, no shake. Grade: warm lifted shadows, punchy mids, controlled highlights, 10% grain, volumetric light, 8K texture on stone, timber and fabric. Sound bed: room tone, soft HVAC hush, one curtain flutter. Final frame: the window wall centred with the terrace beyond, thirds held. No text, no people, no flat wash light, no HDR glow.

[OVERRIDE] soft flare allowed · soft-focus plane · slow pacing
Low-angle golden hour light streaming through a line of trees, backlighting a young man walking a quiet lane with a bicycle, warm rim on his hair and shoulders. Atmospheric haze fills the air; cinematic bloom on the highlights; one gentle sun flare crosses the upper-left. Foreground: out-of-focus wild grass; midground: the subject on the right third at about 30% of frame height; background: melting bokeh of leaves and light. Grade: warm gold, soft green, cream and earthy brown, desaturated, lifted warm blacks, low contrast, compressed airy mids. Camera: Sony A1, 85mm f/1.4 at f/1.6, ISO 100. True skin texture, 10% organic grain, soft not plastic. No harsh sharpness, no midday light, no saturation spikes, no fake glow.

A 35mm analog film still: a woman sits on the edge of an old bed in a dim apartment, looking away toward a dark window, natural slumped posture, unposed. Single motivated practical: a small table lamp at 2900K on the left, dim, underexposed about one stop, hard warm falloff into deep retained shadow across the right of the room. Blacks lifted slightly and tinted warm, never crushed. Aged interior: peeling paint, a worn rug, a chair with clothes over its back, emptiness as emotion. Muted warm palette of brown, amber, faded yellow, olive shadow, dusty beige. Optics soft and shallow, vintage 35mm character, slight vignette. Physical film imperfection: visible grain, faint dust, halation around the lamp, a soft light leak at the frame edge. No brightness, no second light source, no clean modern surfaces, no sharpness, no posing at camera.

Identity strength maximum, style strength medium, from the attached face reference. A peaceful environmental portrait: the subject sits small, about 25% of frame height, on a thick tree branch arcing diagonally over a calm pond, bare feet hanging above the still water, wearing simple minimal light clothing. Time: 3 to 5 PM, low sun 20 degrees, backlight filtering through the leaf canopy as a dappled scrim, warm rim separation on the hair, gentle volumetric haze in the air, around 5000K, warm but not sunset. The water mirrors the branch and figure in a soft reflection. Lens 85mm at f/3.2 so the branch and reflection stay legible; foreground leaves softly out of focus. Grade: warm-green matte film look, lifted blacks, muted greens, protected skin tones, low-medium saturation, 8 to 10% grain, soft bloom. Mood: contemplative. No sunset orange, no wide-open melt of the background, no face morph, no busy wardrobe, no crowds.

MODE: HUMAN. Technical signature suspended: available light only, no added sources, authenticity over control.
A teacher in a small community classroom laughs mid-sentence with two students, caught unscripted from the side, window light from the left as the only source, honest shadows falling naturally. Handheld with soft operator breath, slight reframe as the laugh lands, 0 to 8s continuous. Real skin, real wrinkles, real chalk dust in the light, no beautification. Grade neutral and organic, gentle grain, nothing lifted or stylized. Sound bed: overlapping laughter, a chair scrape, a ceiling fan. No score inside the shot. No staging, no gloss, no two-source lighting, no gimbal smoothness.

MODE: ANALOG. 8K texture rule suspended; imperfection is the aesthetic.
A skater pushes off down a sun-bleached seaside promenade, shot as if on a consumer camcorder: lo-fi grain, chroma bleed on the red of his shirt, faint scan lines, a timestamp-free VHS softness, colours shifted warm and slightly magenta. Whip pan follows the push-off, 0 to 2s; holds a loose tracking frame, 2 to 6s, with authentic camcorder wobble. Highlights clip gently, blacks are milky, edges soft. Sound bed: wheels on concrete, wind hitting a cheap mic, distant radio. No modern sharpness, no stabilization, no cinematic grade, no film-grain overlay pretending to be VHS.

MODE: DARK. Low-key throughout: dark red base, amber-gold escalating glow.
A long corridor in an old house at night, lit only by a dying amber bulb at the far end over a deep red-tinted darkness. A child stands motionless on the right third, midground, back to camera, gaze locked on the far door. 0 to 4s: stillness; the bulb pulses once and the amber glow strengthens slightly. 4 to 8s: the far door drifts open two inches with no visible cause; the glow escalates, spilling gold along the floorboards toward the child. 8 to 10s: the child's head tilts a few degrees; slow 5% push-in the entire runtime. Maximum texture: wood grain, dust, peeling wallpaper, true fabric on the pyjamas, deep 8K detail in shadow. Contrast held high, blacks rich but never empty. Sound bed: house settling, a filament hum, one floorboard note at 8s. Final frame: door ajar on the left third, child silhouetted right, gold pooling between them. No jump-scare imagery, no gore, no music, no text.

The Debug Room

When a generation fails, do not rewrite the whole prompt. Diagnose the starved axis, change one lever, run again. The failure table below covers 90% of bad outputs.

The Global Negative Block

Applied to every output, in every mode, unless deliberately overridden. Negatives are the immune system of the signature: they keep the amateur tells out.

NO: lens flare artifacts · text overlays · plastic skin · fake AI glow · motion blur smear · overexposed sky · anime stylization · cartoon rendering · single flat wash light · middle-ground blur · handheld shake · CGI-obvious renders · teal-orange colour grading extremes · unmotivated VFX · generic stock footage feel · face morph · background bleed · crushed blacks

The Open Brain API

The entire operating system, machine-readable and free. A static, keyless, open-source API you can fetch from any app, agent, or LLM. MIT licensed: build tools on it, ship products with it, teach with it.

Use it in code

const brain = await fetch("https://faimglobal.com/cinematicbrain/brain.json").then(r => r.json());
const dark = brain.modes.find(m => m.name === "DARK");
const move = brain.movement_vault.find(v => v.name === "Orbit clockwise");
console.log(dark.law, move.prompt);   // no key, no signup, CORS-friendly

Use it in any LLM

Fetch brain.txt and paste it as a system prompt. The model inherits KISC, the modes, the axis budget, movement grammar and the negatives in one move. That is the text brain.

Install the Brain as a Claude Skill

The entire operating system, packaged as an installable skill for Claude. One download, one click, and every conversation you have with Claude carries the full Brain: the five modes, the six-axis budget, the Engine Playbook, the Co-Writer Method, character locking, continuity law, the debug discipline, all of it.

The Learning Path

How to teach this system to someone else (or to your future self starting over). Four weeks, one deliverable per week, taste built through constraint.

Each stop tells you the drill to run now, with a copyable exercise. Your position is saved.

Teaching principles

• Teach the four-part unit: every lever is taught as what it controls, the engineering value, the vocabulary, the failure mode.
• Constraint before freedom: closed profiles first; personal style is earned by mastering someone else's system and then breaking one rule with intent.
• The change log is sacred: one dial per generation, every change written down. Taste without records is luck.
• Story outranks everything: if a frame is beautiful and means nothing, it fails the brief. KISC, always.

The Gamified Academy

Reading builds knowledge; playing builds reflexes. The Academy turns this entire brain into a game: a level map you climb, mini-games that drill each skill, quiz bosses that test you, XP that tracks you, and a personalised certificate waiting at the end.

Learn with Joby

This brain is the self-serve version of what I teach in person. If you or your team want the guided version, with live feedback, real briefs and my eyes on your frames, this is what I offer.

Send a Message

Write it here, send it straight to my WhatsApp. No forms, no waiting, no middleman.

Connect with Joby

This brain is written and maintained by Joby Thuruthel, AI filmmaker and director, co-founder of FAIM, an AI-powered creative agency in Bahrain serving the GCC and beyond. For films, training, collaborations or questions about anything in this system, reach out directly.

If this brain helped you make something, send it to me on WhatsApp or tag @joby_thuruthel. Seeing the system in other people's hands is the whole point.