All guides

Genres 14 min August 2026

AI Video by Genre

Genre isn't a separate craft — it's the same set of production dials, set differently. A comedy cut at thriller pace isn't a slow comedy, it's a film that doesn't work.

Japesh Singhal

Craft & production

Download skill

Every genre answers the same short list of questions — how long is a shot, what does the camera do, how is it lit, what carries the meaning when nobody's talking — just with different values. Set these dials deliberately before shot planning. Anything left unset gets the model's default, and the model's default is identical regardless of genre — which is exactly why unset films all look alike.

The dials, in one list
Average shot length, shot-size range, camera movement, lens choice, lighting, color palette, sound design, music role, performance register, cut style, and dialogue density. Every genre subpage below sets these deliberately for its genre, then adds the one paragraph of judgment a table of values can't express.

A comedy ad, a thriller trailer, a documentary-style product film — resolve a hybrid this way: the story shape and the dials both come from the primary genre; the secondary genre gets exactly one dial, as a deliberate accent, not a second full set. A comedy ad is comedy dials with one commercial dial (the ending demonstrates rather than jokes). A thriller trailer is thriller dials with one action dial (the final cut rate).

The most common cause of tonal mush
Two full sets of dials in one film. Pick a primary genre and commit to it; borrow exactly one thing from the secondary, never a whole register.

Almost every film has one moment where the dials change — the joke lands, the threat becomes real, the product arrives. Design that shift explicitly rather than letting it happen by accident:

From → ToWhat changes
Comedy → sinceritySlow the average shot length, drop the music out, cut to a single
Calm → threatShorten the lens, kill the ambience, hold the shot too long
Problem → solutionBring the palette back, lift the light, let the music resolve

If the dials never change across a film, it has no turn — and a film with no turn is a loop of the same idea at the same intensity from start to finish.

Set and commit, before writing a single shot

Platform tier (decided before this list) — Primary genre + secondary accent (max one dial) — Story shape + target shot length — Shot-size range + camera movement vocabulary — Lens set — Light (source, direction, ratio) — Palette (3-4 colors) — Sound design density — Music role per section — Performance register — Transition grammar — Where the turn is, and which dials move.

01

Most real briefs are hybrids

A comedy ad, a thriller trailer, a documentary-style product film — resolve a hybrid this way: the story shape and the dials both come from the primary genre; the secondary genre gets exactly one dial, as a deliberate accent, not a second full set. A comedy ad is comedy dials with one commercial dial (the ending demonstrates rather than jokes). A thriller trailer is thriller dials with one action dial (the final cut rate).

The most common cause of tonal mush
Two full sets of dials in one film. Pick a primary genre and commit to it; borrow exactly one thing from the secondary, never a whole register.
02

The turn

Almost every film has one moment where the dials change — the joke lands, the threat becomes real, the product arrives. Design that shift explicitly rather than letting it happen by accident:

From → ToWhat changes
Comedy → sinceritySlow the average shot length, drop the music out, cut to a single
Calm → threatShorten the lens, kill the ambience, hold the shot too long
Problem → solutionBring the palette back, lift the light, let the music resolve

If the dials never change across a film, it has no turn — and a film with no turn is a loop of the same idea at the same intensity from start to finish.

03

A checklist before shot planning

Set and commit, before writing a single shot

Platform tier (decided before this list) — Primary genre + secondary accent (max one dial) — Story shape + target shot length — Shot-size range + camera movement vocabulary — Lens set — Light (source, direction, ratio) — Palette (3-4 colors) — Sound design density — Music role per section — Performance register — Transition grammar — Where the turn is, and which dials move.

04

AI video in the comedy genre

DialComedy setting
Shot length1.5-3s
Shot rangeMediums; close-up saved for the punchline
MovementLocked, or a slow push
LensWide for chaos, normal for the exchange
LightEven, warm, high key
PaletteWarm, saturated
Sound designClean, dry, ambience-led
MusicSparse; out on the joke
RegisterPlayed straight
CutsHard, on the beat
Dialogue densityHigh, short lines

What it lives or dies on: rhythm. The joke is timing, and timing is the edit — land the cut one beat early rather than one late. Everything else can be adequate if the rhythm is right; nothing saves a comedy that drags. The reaction shot is the joke, so coverage of the listener isn't optional. See comedy timing for the full beat architecture.

05

AI video in the drama genre

DialDrama setting
Shot length4-8s
Shot rangeSingles
MovementSubtle push
LensNormal to long
LightSoft ratio, directional
PaletteDesaturated, narrow
Sound designIntimate, sparse
MusicMotif, late entry
RegisterSuppression
CutsDissolves can earn their place
Dialogue densityMedium, subtext

What it lives or dies on: the withheld. Meaning lives in what a character does not say and does not do — longer takes, fewer cuts, performances built on suppression. The temptation to cut away from discomfort is the thing to resist.

06

AI video in the thriller genre

DialThriller setting
Shot length3-6s, then sudden
Shot rangeTight; withhold the wide
MovementSlow creep; static holds
LensLong, compressed
LightHard, single source
PaletteCold, limited
Sound designDesign-forward; the room breathes
MusicSustained tension bed
RegisterContainment; one break
CutsHold, then hard
Dialogue densityLow

What it lives or dies on: information control. The audience must know slightly less, or slightly more, than the character — decide which, deliberately, per scene. Withhold the wide shot; a viewer who can see the whole room isn't tense. Sound carries more than picture.

07

AI video in the action genre

DialAction setting
Shot length0.8-2s
Shot rangeFull range plus inserts
MovementHandheld, whip, tracking
LensWide and long, mixed
LightHigh contrast, motivated
PaletteContrast-driven
Sound designImpact-led, layered
MusicPercussive, driving
RegisterEconomy, intention
CutsCut on impact
Dialogue densityVery low

What it lives or dies on: geography and consequence. Fast cutting only works if the audience knows where everyone is — otherwise it's noise. Establish the space, then break it. Cut on impact, and give every hit a reaction shot after it; action with no consequence reads as choreography.

08

AI video in the horror genre

DialHorror setting
Shot lengthLong holds, then a snap
Shot rangeNegative space; partial framing
MovementSlow drift
LensWide, slightly distorted
LightUnderexposed, one edge
PaletteMonochrome plus one color
Sound designSilence as a weapon
MusicAbsent, then a stab
RegisterDelayed comprehension
CutsHold past comfort
Dialogue densityVery low

What it lives or dies on: duration. Fear is generated by holding a shot past the point of comfort, not by the reveal — the cut you don't make is the technique. Silence does more than any sting, and a stab only works after real silence.

09

AI video in the romance genre

DialRomance setting
Shot length3-5s
Shot rangeWide for scale, close-up for humanity
MovementSlow push, slight float
LensLong, shallow
LightSoft, warm, backlit
PaletteWarm, soft
Sound designBreath and cloth
MusicMelodic, patient
RegisterMicro-behavior
CutsHold on faces
Dialogue densityLow

What it lives or dies on: proximity. The story is the distance between two people, measured in frames — hands, breath, the moment before contact. Cut on the almost, not the event.

10

AI video in the sci-fi genre

DialSci-fi setting
Shot length3-5s
Shot rangeWide for scale
MovementSlow, weightless, precise
LensWide for scale
LightCool, hard, practical-heavy
PaletteCool, engineered
Sound designDesigned, synthetic, tonal
MusicAtmospheric drone
RegisterRestraint amid scale
CutsHold on scale
Dialogue densityMedium, expository risk

What it lives or dies on: consistency of rules. The world must obey its own logic without explaining it. The failure mode is exposition — characters telling each other things they both already know. Show the rule working; never state it.

11

AI video in the documentary / testimonial genre

DialDocumentary setting
Shot length5-10s
Shot rangeFixed frame, one size
MovementLocked off
LensNormal
LightFlat, honest
PaletteNeutral
Sound designUntreated realism
MusicNone or minimal
RegisterImperfect, real
CutsMinimal
Dialogue densityHigh

What it lives or dies on: imperfection, but not on its own. Pauses, restarts, looking away to think, an answer that doesn't quite answer — anything too smooth reads as scripted, which destroys the only thing the form has.

On hard-exit platforms, raw needs a scaffold
Fully raw, unstructured footage measurably underperforms short-form. The pattern that wins there is closer to "mockumentary" — a raw aesthetic paired with a real narrative hook up front, not imperfection alone. On long-form (YouTube, LinkedIn) raw can stand on its own longer without that scaffold. (Reported as practitioner consensus, not a controlled study.)
12

AI video in the commercial genre

DialCommercial setting
Shot length1.5-3s
Shot rangeFull range
MovementMotivated, smooth
LensNormal
LightClean, flattering
PaletteBrand-led
Sound designPolished
MusicHook-led
RegisterNaturalism, low
CutsHard, rhythmic
Dialogue densityLow

What it lives or dies on: one idea, demonstrated. Thirty seconds holds exactly one idea. The strongest ads show the product doing its job rather than claiming it. Match the genre dials of whatever the ad is pretending to be (a comedy ad, a documentary-style ad), then cut it tighter than that genre would on its own.

13

AI video for weddings and live events

DialWedding/event setting
Shot lengthBimodal — uncut through the ritual, 1.5-4s in montage
Shot rangeWides for venue/scale, singles for reaction, macro detail in montage
MovementSlow-motion as a signature; gimbal glide, drone establishing
LensShallow, 50-85mm for portraits/detail; wide for the venue
LightNatural/available; golden-hour is a genre beat
PaletteWarm, filmic, soft highlights
Sound designLive audio ducked (not silenced) under vows/speeches; music-led elsewhere
MusicBimodal — narrative (real audio drives structure during ceremony) and editorial (music leads during reception)
RegisterSincere, sentimental, idealized — never comedic, never raw
CutsHard, beat-synced in montage; uncut through ritual
Dialogue densityLow in trailer form; high (full vows/speeches) in the full-length film

What it lives or dies on: the unrepeatable moment. There's no take two, and the craft protects two things at once. First, the couple's actual voice — duck music under vows and speeches, never replace it. Second, the emotional shape a real event doesn't hand you for free, which is what the music-led edit is for. Real practice runs two modes, not one: narrative (the ceremony, where real audio drives structure) and editorial (reception and highlights, where music and rhythm lead) — the strongest work blends both rather than picking one. Slow motion isn't decoration here — it's the genre's way of saying this mattered.

14

AI video for music videos

DialMusic video setting
Shot lengthFast, BPM-locked, 0.5-2s
Shot rangeFull range, biased to performance medium/close
MovementHighly kinetic — whip pans, speed ramps, in-camera transitions
LensWide range, stylized/experimental is legitimate
LightBold, saturated, practical-heavy; can break naturalism
PaletteBold/saturated or high-contrast mono — extremity is licensed here
Sound designSubordinate to the track; SFX only as transition punctuation
MusicNot a layer — the structure. Every cut answers to it
RegisterPerformative — performance of a song, not behavior
CutsLocked to BPM: 60 ÷ BPM = seconds per beat
Dialogue densityNone — lyrics are the only "dialogue," lip-synced

What it lives or dies on: the beat is the director. Every other genre asks the edit to serve character or story; this one asks it to serve tempo. This is a real, tooled technique, not just a feel — 60 ÷ BPM = seconds per beat, and most professional editing software ships dedicated beat-detection features for cutting to it. A cut that lands off the beat reads as a mistake even when nothing else is wrong.

15

AI video for product explainers

DialProduct/explainer setting
Shot lengthBimodal — 1-3s feature cuts, 3-8s "watch it work" holds
Shot rangeOften no camera at all — motion graphics/screen capture
MovementAnimation movement (element transitions, virtual camera)
LensNot applicable to live-action; virtual-camera focal choices used sparingly
LightClean, flat, brand-consistent — no cinematographic design
PaletteStrict brand palette, not mood-driven
Sound designMinimal, functional (clicks, soft whooshes)
MusicBed only, ducked under voiceover always, never load-bearing
RegisterHelpful, clear, teacher-authoritative — not theatrical
CutsHard, feature-to-feature, governed by the voiceover script's beats
Dialogue densityHighest of any genre — script-first, roughly 150 words per minute of finished video

What it lives or dies on: the script is the storyboard. Write the voiceover first, at roughly 150 words per minute of finished video (a 90-second explainer script runs about 225 words) — every visual exists to land one sentence of it. The moment the visuals get impressive enough to outrun the script, the video stops explaining and starts decorating, and the viewer stops understanding the product.