How to Write Cinematic AI Video Prompts: A Filmmaker's Complete Guide

Most people write AI video prompts like they're texting. "Show me a man running through a city at night." And they get exactly that — technically accurate, cinematically dead.

The filmmakers producing jaw-dropping AI footage are doing something different. They're not just describing scenes. They're directing them. This guide teaches you that difference.

Why Cinematographic Language Works

AI video models — especially Runway Gen-3 and Google Veo 2 — were trained on vast libraries of professionally labeled footage. That training data used real cinematography terminology. When you write "shallow depth of field on an 85mm lens," the model doesn't have to interpret your intent. It knows exactly what that looks like.

Vague language forces the model to make decisions for you. Precise language lets you make those decisions yourself.

The 6-Part Cinematic Prompt Framework

Every great AI video prompt answers six questions. You don't always need all six, but you should always know all six.

1. Shot Type — What's the Frame?

This is the single most impactful element of your prompt. Shot type determines the entire emotional register of the frame.

| Shot Type | Code | When to Use | |-----------|------|-------------| | Extreme Close-Up | ECU | Emotion, detail, intimacy | | Close-Up | CU | Character focus, reaction | | Medium Close-Up | MCU | Dialogue, connection | | Medium Shot | MS | Action, relationship | | Medium Wide | MW | Context + character | | Wide Shot | WS | Environment, isolation | | Extreme Wide | EWS | Scale, epic, loneliness |

Example: Medium close-up on a woman's face as she reads a letter vs. Extreme wide shot — a lone figure dwarfed by the ruins of a flooded city. Same subject matter, completely different emotional impact.

2. Lens — How Does It See?

Lens choice is where most beginners leave massive quality on the table.

  • 24–35mm (wide): Environmental, slightly distorted, immersive. Great for action, environments, unease.
  • 50mm (standard): Closest to human eye perception. Naturalistic, documentary.
  • 85mm (portrait): Slight compression, beautiful subject isolation. Classic cinematic dialogue and character work.
  • 135mm+ (telephoto): Heavy compression, subjects pop from backgrounds. Surveillance feeling, distance.
  • Anamorphic: Oval bokeh, horizontal lens flares, that widescreen cinematic quality. Use whenever you want the "movie" look.

Practical prompt: 85mm anamorphic lens, shallow depth of field, subject sharp against a blurred neon cityscape

3. Lighting — The Soul of the Frame

Roger Deakins doesn't light a scene — he sculpts emotion with light. Your prompts should too.

Lighting setups to know:

  • Motivated lighting: Light comes from sources visible in frame (lamps, windows, screens). Feels authentic.
  • Chiaroscuro: Deep shadows, dramatic contrast. Think Caravaggio paintings or noir.
  • Golden hour: Warm, low-angle natural light. Everything looks beautiful.
  • Practical-heavy: The scene is lit almost entirely by in-world sources — creates intimacy and grit.
  • High-key: Bright, even, minimal shadows. Commercial, optimistic.
  • Low-key: Dark overall with selective illumination. Tension, mystery, drama.

Prompt example: Warm amber practicals raking light across the character's face from below, deep shadow filling the right side of frame, cold blue spill light from an unseen window in the background

4. Camera Movement — Does It Breathe?

Static cameras feel formal and controlled. Moving cameras feel alive. Know your vocabulary:

  • Static: Locked off, controlled, formal
  • Handheld: Organic, present, slightly unstable — implies documentary or intimacy
  • Dolly in: Building intensity, growing importance
  • Dolly out / Pull back: Revelation, isolation, loss of intimacy
  • Pan: Survey, reveal, connection between two elements
  • Tilt: Power dynamics (tilt up = subject gains power)
  • Crane/Jib rising: God's-eye view, scale, epic departure
  • Tracking shot: Following action, momentum

5. Atmosphere & Environment

Atmosphere transforms footage from recording to cinema:

  • Rain-slicked streets with neon reflections
  • Dust motes floating in shafts of late afternoon light
  • A thin haze diffusing the background into soft impressionism
  • Heat shimmer distorting the horizon
  • Fog rolling at ankle height through ancient forest

6. Emotional Tone

End your prompt with the feeling you want. This guides the model's stylistic decisions for everything it didn't already specify:

  • The frame should feel like a memory, slightly dreamlike
  • Tense, suffocating, nowhere to run
  • Melancholic beauty — sad but gorgeous
  • Kinetic, breathless, velocity

Putting It All Together: Before & After

Bad prompt:

A scientist works in a lab at night.

Great prompt:

Medium close-up — a scientist's weathered hands examine a glowing vial in a darkened lab, 85mm lens with shallow depth of field, the background a blur of blue-lit equipment. The only light is the cool bioluminescent glow from the vial catching the angles of her tired face. Static camera, perfectly still, like the world is holding its breath. Atmosphere: quiet dread building beneath professional composure.

Same scene. Completely different footage.

Common Mistakes to Avoid

1. Over-describing the story, under-describing the frame AI models are cameras, not storytellers. Tell them what the camera sees, not what the narrative means.

2. Conflicting instructions "Handheld close-up wide shot with dramatic crane movement" confuses the model. Pick one approach.

3. Ignoring time of day Light changes everything. Always specify: golden hour, blue hour, midnight, harsh midday, overcast grey.

4. Forgetting lens flares and atmosphere These small details push footage from "rendered" to "cinematic." Subtle anamorphic lens flares at light sources costs you four words and transforms the image.

The Shortcut: Let ShotForge Handle the Framework

Learning all of this takes time. ShotForge automates the framework — you choose your shot type, lens, lighting mood, camera movement, and AI tool, and it assembles the cinematographic language into a professional prompt instantly.

Generate your first cinema-grade prompt →

The best cinematographers aren't the ones who break the rules — they're the ones who know every rule so well that breaking them is a choice, not an accident. The same applies to AI direction.