AI Image & Video Prompt Vocabulary

✦ How a strong prompt is built

Every model reads a prompt roughly front-to-back and weighs earlier words more heavily. Order is part of the vocabulary. Default to this structure:

Worked example

A weathered fisherman mends a net on a wooden dock (subject & action) — medium shot, eye-level, slow push-in (camera) — soft golden-hour backlight (lighting) — warm desaturated tones, Kodak Portra grade (color) — quiet, nostalgic (mood) — shot on 35mm film (technical)

1 Concept & Creative Direction

Before touching camera or lighting language, get the idea itself into words the model can act on.

• Logline — the concept in one sentence: "A courier discovers her package is alive." Write this first, always.
• Tone words — 2-3 adjectives that anchor the whole piece: gritty, whimsical, aspirational, unsettling, nostalgic, clinical, romantic, kinetic.
• Reference language — "in the style of Wes Anderson," "evokes 1970s National Geographic photography." Naming a known aesthetic does more work than ten adjectives.
• Narrative beat words — establish, escalate, reveal, collide, resolve. Useful for structuring a shot sequence or storyboard.
• Genre framing — commercial, editorial, documentary, music video, product hero, UGC/influencer, cinematic trailer. State this early; it changes how the model weights realism vs. stylization.

2 Storytelling & Pacing

Vocabulary and camera choices only pay off if the sequence they sit in is paced deliberately. This matters even for a single AI-generated shot — every shot implies a "before" and "after" — and it matters far more once you're stringing shots together or using a model's native multi-shot mode.

Narrative structure vocabulary

• Logline — the story in one sentence (see Concept & Direction). Write this before anything else — it's the north star for every downstream shot.
• Three-act structure — setup → confrontation (rising action) → resolution (climax, then a brief denouement). Useful even for a 15-second ad: setup (2-3s) → tension (majority of runtime) → payoff (final beat).
• Inciting incident / hook — the moment that breaks the status quo. In short-form this needs to land in the first 1-2 seconds or the viewer scrolls past.
• Rising action — each beat raises stakes, tempo, or information. Describe it explicitly: "beat 2 escalates beat 1."
• Climax — the peak of tension or payoff — biggest camera move, tightest framing, or loudest visual contrast usually lives here.
• Denouement / resolution — the settling beat after the climax; often a wider, calmer, slower shot to land on.
• Character arc — how a subject's state changes from first shot to last. Even product films benefit from an implied arc: problem → relief.
• Beat sheet — a shot-by-shot outline written before prompting. Do this for anything longer than 2-3 shots — cheaper to fix pacing on paper than after ten generations.

Pacing & rhythm vocabulary

• Cutting rate — how often shots change. Fast-cut reads urgent / high-energy; long-take reads contemplative or prestige.
• Montage — a rapid sequence compressing time or building an idea. Describe the throughline: "morning routines, each shot 1-2s, cutting on movement."
• Cut on action — cutting mid-movement (a hand reaching, a door swinging) to make cuts feel invisible and kinetic.
• Cross-cutting / parallel editing — alternating between two storylines happening at once, to build tension or comparison.
• J-cut / L-cut — audio from the next/previous shot starts before/after the picture cuts. "Sound leads picture by half a second."
• Breathing room — a deliberately quiet, wider, slower shot after an intense sequence — resets attention.
• Tempo ramp — cutting rate accelerates toward the climax, then resets to a single long resolution shot.

Transition vocabulary

• Hard cut — an instant change of shot. The default; use for pace, not punctuation.
• Cross-dissolve — one shot fades into the next. Signals time passing or a dreamlike connection.
• Match cut — two shots joined by a shared shape or motion (a spinning wheel to a spinning planet). Powerful for AI multi-shot — describe the shared element explicitly.
• Smash cut — an abrupt, jarring cut, often quiet-to-loud, for shock or comedy.
• Whip-pan transition — a fast pan blurs into the next scene — very AI-video-native since it hides continuity limits.
• Morph transition — one subject fluidly warps into the next. Distinctly AI-native: "the frame seamlessly morphs from [A] into [B]."
• Fade to black/white — signals a hard ending or chapter break.

Directing pacing inside AI video tools

• Native multi-shot models — Kling can generate up to 6 shots from one prompt holding continuity — describe the beat sheet directly in the prompt.
• Timeline prompting — Seedance supports timestamps ("0-2s: wide establishing, static — 2-4s: push-in") so pacing is structural, not just prose.
• Beats within one shot — for single-shot models keep to 2-3 sequential beats max ("she reaches, hesitates, then pulls it open").
• Edit separately-generated shots — generate each shot individually and assemble the pacing yourself for full control.
• State the pace explicitly — "quick, punchy cuts," "a single unbroken take," "building tempo" are all valid prompt language.

3 Composition & Framing

Shot size ladder — the same subject, the same distance, cropped differently. The single most useful lever for controlling how much information is in a shot.

Reference photos generated with Nano Banana Pro for this guide.

What each abbreviation means

Use the full term or the abbreviation — both are widely understood. Ordered from tightest to widest:

• ECU — Extreme close-up — a single detail fills the frame: an eye, a hand, a logo. Tension, product detail, emotional intensity.
• CU — Close-up — head and shoulders. The default for emotional connection, dialogue, portraiture.
• MCU — Medium close-up — chest to top of head. Common for interviews and talking-head shots.
• MS — Medium shot — waist up. The workhorse conversational framing.
• Cowboy shot — mid-thigh up. Common in fashion and lifestyle — shows an outfit while keeping the face prominent.
• MLS — Medium long shot — knees up. Transitional between a cowboy shot and a full wide.
• LS / WS — Long / wide shot — full body plus environment. Establishes where the subject is.
• EWS — Extreme wide shot — the subject is small within a vast environment. Scale, isolation, grandeur.

Framing techniques

• Rule of thirds — subject off-center on a grid intersection
• Leading lines — roads, railings, architecture that pull the eye to the subject
• Negative space — large empty areas for breathing room
• Symmetry / centered composition — formal, deliberate; Wes Anderson / brutalist aesthetics
• Frame within a frame — doorway, window, mirror boxing in the subject
• Foreground layering — an out-of-focus foreground element adds depth
• Depth of field — shallow (sharp subject, blurred background) vs. deep (all in focus)

4 Camera Angle Vocabulary

Reference photographs of the same subject shot from each angle — each card notes the emotional effect it creates.

Reference photos generated with Nano Banana Pro for this guide.

5 Camera Movement Vocabulary

Where AI video genuinely separates from stills — the vocabulary that makes a shot feel directed rather than default.

Zoom vs. dolly: a zoom changes magnification only — background compresses and flattens. A dolly/push-in physically moves the camera — perspective shifts and the background "wraps" around the subject. State which one you mean; models will otherwise guess.

See it in motion

Fifteen short clips, each isolating one movement — play them side by side and describe out loud what the camera is doing before you check the caption.

Aerial/drone, pan, tracking, and rotating clips sourced from Mixkit (free stock video license). The rest generated with Seedance 2.0 for precise, correctly-labeled camera motion.

6 Lighting Vocabulary

Real portrait references for each lighting concept — same subject, same background, only the light changes.

Quality

Classic setups

Chiaroscuro — the umbrella term for all of these: extreme light/dark contrast, painterly.

High-key vs. low-key

Natural light by time of day

• Golden hour — warm, low-angle sun near sunrise/sunset; soft and flattering
• Blue hour — cool, deep-blue ambient light just before sunrise/after sunset
• Overcast / softbox sky — even, shadowless, flattering diffuse light
• Harsh midday sun — high contrast, hard shadows, bleached highlights
• Moonlight — cool, low-level, often stylized blue in film

Artificial & practical light

• Neon-lit — saturated colored light, urban/cyberpunk
• Tungsten / warm practicals — orange-toned interior bulbs, cozy or retro
• Fluorescent — cool, flat, slightly green — clinical, institutional
• Candlelight / firelight — warm, flickering, low, intimate
• Practical lights in frame — lamps, signage, screens motivating the scene's light

See it in practice

Reference photos generated with Nano Banana Pro for this guide.

7 Color & Grading Vocabulary

Palette structure

Grading styles

8 Style & Medium Vocabulary

Lens & film references

• Shot on 35mm film — natural grain, filmic color response
• Anamorphic lens — horizontal lens flares, oval bokeh, widescreen cinematic look
• Fisheye — extreme wide-angle distortion, curved edges
• Macro lens — extreme close focus
• Telephoto compression — flattens depth, isolates subject from a distance
• Shallow depth of field / bokeh — soft, blurred background circles of light
• Film grain — textured, organic noise — vintage authenticity

Aspect ratio

16:9 (widescreen/YouTube) · 9:16 (Reels/TikTok/Shorts) · 1:1 (feed/product) · 2.39:1 (cinematic anamorphic) · 4:5 (portrait feed)

9 Props, Set Design & Environment

• Set dressing — background objects that build a world (props, signage, clutter or minimalism)
• Hero prop — the one object the shot must feature; describe material/color/brand fully
• Practical props — background objects that don't need to read clearly, described loosely
• Texture & material call-outs — "brushed steel," "worn leather," "cracked plaster," "condensation on glass"
• Environment descriptors — interior/exterior, weather (light drizzle, dense fog, heat haze), scale (cavernous, cramped)
• Wardrobe / costume lock — describe a recurring character's outfit once, then state it "must remain exactly identical across all shots unless a change is listed."

10 Mood & Atmosphere Vocabulary

Pick words that reinforce, not contradict, your lighting and color choices. Pair the mood word with a physical cause: not just "eerie," but "eerie — fog pooling at ankle height, single flickering light."

11 Consistency & Control Techniques

AI models forget. These claw back control.

• Explicit consistency statements — "keep [character/product] visually identical across all shots"; name subjects — a named character holds together far better than "a man."
• Negative prompts — state what should not appear (extra limbs, text artifacts, watermark, blurry).
• Seeds — a fixed seed number locks the model's starting randomness, letting you tweak the prompt while keeping composition stable.
• Reference / image prompting — upload a reference image so the model matches composition, character, or style.
• Start/end frame interpolation (video) — supply a start and end frame, instruct "smoothly interpolate, maintain character consistency."
• Still-image rule for storyboards — every panel must be something a photographer could capture in one frozen frame. "She nods" isn't valid; "she looks at him, chin lowered, small relieved expression" is.

12 Tool-Specific Cheat Sheet

Every tool below reads this guide's shared vocabulary — but each has its own syntax, parameters, and quirks. AI tools move fast; treat this as a living reference and check each platform's own docs before a big client deliverable.

Across every tool: put subject, action, and camera in one clear block, then lighting/color/style after. Front-loading what should logically come first produces cleaner results.

🛠 Interactive Prompt Builder

Assemble a prompt using the formula from this guide. Change any field and the preview updates live.

📖 Quick-Reference Glossary

🎴 Flashcard Practice Drill

Run through the glossary as flashcards. Click the card to flip.