Orelon logoOrelon
料金

Cinematic Techniques for Vertical Video That Actually Read

2026年11月13日 · Orelon Team 著

AI動画テンプレートを見る

着想のためにコミュニティ作品をいくつか閲覧し、任意のテンプレートを開いて Orelon で作成を続けましょう。

Composition, lighting, movement and depth-of-field techniques that make vertical short-form video feel cinematic, plus a repeatable AI generation workflow.

Most vertical shorts do not fail because the idea is weak. They fail because the frame is flat, the light is even, and the camera sits still in a way that feels accidental rather than deliberate. Cinematic quality in a 9:16 frame is not about owning a cinema camera. It comes from four decisions repeated in every shot: what sits inside the frame, how the camera moves, where the light comes from, and how much of the image is allowed to go soft. Get those four right and a phone clip reads like a film. Get them wrong and expensive footage drifts toward amateur.

This guide treats those decisions as a working system. It covers how they apply to a tall frame, how to describe them precisely enough for a generative model to obey, and how to keep generated footage and shot footage looking like they belong to the same piece.

A cinematic look is a constraint set, not a filter

Consistency is what audiences read as quality. They usually cannot explain why one creator's clips feel expensive while another's feel improvised, but the cause is almost always the same: the first creator repeats a small number of rules, the second makes each shot independently.

Define four rules before you shoot or generate anything, and hold them for the whole series.

  • One dominant light direction. A hard side key in one shot and flat frontal light in the next breaks the illusion instantly, even if both frames are individually pleasing.
  • One lens family. Wide, normal, and long optics each bend space differently. Rotating between them at random makes a sequence feel assembled rather than directed.
  • One palette rule. Pick a shadow tint and a highlight tint and keep them. Five shots graded in five directions look like five different videos stitched together.
  • One movement logic. If the camera pushes in during dialogue, it should not suddenly orbit during the reaction shot.

Write those rules into a single page you can keep open on a second monitor. A usable look brief reads like this:

  • Palette: cool shadows, warm skin tones, muted greens in the environment.
  • Light: single soft key from camera left, low ambient, rim light on the subject's right shoulder.
  • Optics: long-lens compression, background soft but still readable, minimal edge distortion.
  • Movement: slow push-ins only, no handheld, no orbits.

That page is more valuable than any individual shot list, because it can be handed to a collaborator, pasted into a prompt, or checked against a finished cut in under a minute.

Composition inside a tall frame

Vertical composition has different physics from landscape. The eye travels up and down a 9:16 frame faster than it travels left to right, which means the upper third carries far more visual weight than it does in a 16:9 image. Techniques designed for widescreen do not fail in vertical video, but they need translation.

Put the eyeline high and keep the headroom honest

Place your subject's eyes on or near the upper horizontal third line. That one habit fixes more vertical shots than any other single change. Then decide what occupies the remaining space. Three reliable patterns:

  • A vertical stack. Subject in the lower third, texture or sky in the upper third, and something in the middle that connects the two. Strong for portraits and product reveals.
  • A dominant diagonal. Subject at the lower-left intersection, light source or horizon at the upper-right. Diagonals read as movement even when nothing in the frame moves.
  • Centered symmetry, used sparingly. Dead-center framing is powerful once per sequence. Used five times in a row, the edit becomes monotonous.

Headroom is where beginners lose the frame. A hand's width of empty space above the head reads as a mistake, not a choice. If you have it, raise the camera or crop in. The exception is deliberate space reserved for captions, which in vertical video usually occupy the upper and lower margins anyway. Plan the text zone before you compose, not after.

Eyeline direction is a second, quieter decision. A subject looking toward the top of the frame reads as aspirational or expectant. A subject looking down reads as introspection or defeat. Neither is accidental once you start noticing it in your own work.

Build three layers in every setup

Widescreen cinematography loves layered depth. Vertical video needs it more, because the narrow frame removes the horizontal escape route the eye would normally take. Without depth, a tall shot collapses into a flat rectangle with a face pasted on it.

Build three layers into every setup:

  1. Foreground. Something close to the lens, slightly soft, occupying ten to twenty-five percent of the frame. A doorframe, a plant stem, a shoulder, the edge of a curtain. It creates intimacy and implies a vantage point.
  2. Midground. Your subject, sharp and fully lit, doing the actual work of the shot.
  3. Background. Environmental context, deliberately lower in contrast than the subject so it does not compete for attention.

When prompting a generative model, spell the layers out. A phrase like blurred leaves close to the lens at the left edge, subject sharp in the midground, distant city lights soft behind gives the model a depth structure to solve. Without it, you get a cutout on a backdrop, and the eye registers the fakeness before the brain can name it.

Use negative space as a design element

Empty space in a vertical frame is not wasted space, it is pacing. A shot with generous negative space above a subject gives the viewer a moment to breathe and gives your editor somewhere to place a title. A shot packed to every edge feels breathless and slightly claustrophobic.

The practical rule: every frame should have one obvious subject and one obvious resting place for the eye. If you cannot point at the resting place, the composition is working against the story.

Camera movement that survives a small screen

Movement is the fastest way to make a shot feel expensive, and the fastest way to make it feel nauseating. On a phone-sized screen, large movements become visual noise. Small, motivated movements read as craft.

Five moves worth mastering

  • Slow push in. Ten to fifteen percent of the frame over three seconds. The most reliable tension builder in short-form video, and one of the easiest instructions for a generative model to follow.
  • Controlled tilt. Start on a detail, end on the face. Vertical frames are built for tilts because the long axis is the interesting one.
  • Lateral track. A short sideways slide that introduces parallax between foreground and background. Even a few centimeters of travel adds three-dimensionality.
  • Handheld drift. Slight, organic float. Excellent for documentary and lifestyle content, disastrous for product beauty shots.
  • Static frame with a moving subject. The most cinematic option and the most underrated. Lock the camera, let a person or vehicle cross the frame, and let composition do the work.

Match the move to the emotional job

Ask what the shot is doing dramatically before you decide how it moves. A push in creates intimacy or dread. A pull out creates isolation or revelation. An orbit creates awe or transformation. A drift creates immediacy. If you cannot name the emotional job in one sentence, do not move the camera. A locked-off frame is never wrong, it is just less interesting than a motivated move and far better than an unmotivated one.

Describe the change, not the equipment

Models respond far better to physical descriptions than to camera jargon. Instead of cinematic dolly, write the camera moves slowly forward toward the subject, the background compressing behind them, shallow focus maintained throughout. That phrasing describes the image change rather than the gear, and generated footage tends to obey it. The prompt library is a useful place to study how movement language is structured before writing your own.

Lighting decisions that create dimension

Most creators lose the cinematic feel here, not because they lack equipment but because they light everything evenly. Even lighting is how you photograph evidence. Directional lighting is how you tell a story.

Key, fill, and rim as three jobs

You do not need three stands. You need three jobs performed.

  • Key. The main source, placed roughly thirty to forty-five degrees off the subject's nose and slightly above eye level. This creates the shadow side that makes a face look three-dimensional.
  • Fill. A weaker source or a bounce card on the opposite side. The goal is not to erase shadows but to control their density. A two-to-one ratio between key and fill looks natural. A five-to-one ratio looks dramatic.
  • Rim or kicker. A light behind and to the side of the subject that separates them from the background. In vertical video this matters enormously, because the frame is narrow and subjects often sit close to the background plane.

Three quick wins that cost nothing:

  1. Shoot near a window, but perpendicular to it. Facing a window flattens the face. Standing at ninety degrees gives you a natural key with falloff across the frame.
  2. Dim the practicals, then power the key. Rooms look cinematic when ambient light is low and one motivated source does the defining.
  3. Use negative fill. A dark jacket, a black card, or simply a wall on the shadow side raises contrast without adding a single fixture.

Use color temperature as a storytelling signal

Mixing temperatures is one of the cheapest ways to look expensive. A warm key against a cool background reads as an evening interior. A cool key against a warm practical reads as a night exterior. The contrast itself creates depth.

Pick a reference — daylight around 5600K, tungsten around 3200K — then deliberately place your subject slightly off the ambient temperature. If the room is cool, make the face warm. If the room is warm, make the face neutral-cool. This single habit separates professional-looking images from snapshots more reliably than any preset.

Depth of field and lens character

Shallow focus is the most recognizable cinematic cue in the vocabulary. It works because it mimics how attention behaves: one plane is sharp, everything else gives way.

The trade-off behind shallow focus

You do not need to memorize aperture numbers to use this. Understand the trade instead. Wider apertures blur the background more but make focus harder to hold, especially on a moving subject in a narrow frame. Narrower apertures keep more of the scene readable but flatten the separation between subject and environment.

For vertical shorts, the useful range keeps the face sharp and the background recognizable but soft. Obliterate the background completely and you lose the context that made the shot interesting. Keep it fully sharp and you lose subject separation. The middle is where most good vertical work lives.

Bokeh as composition

When background highlights blur into soft shapes, they become texture. This is why night scenes with distant point lights feel so cinematic: the bokeh becomes a pattern inside the frame that echoes your composition. Place those highlights in the empty area of your thirds grid and the frame gains structure without adding a single object.

Lens character as narrative choice

A wide lens exaggerates the space around a subject and makes the environment feel present. A long lens compresses the background into the subject and makes the world feel intimate or oppressive. In vertical video, wide lenses are harder because they introduce distracting edge distortion at the top and bottom of the frame. Long lenses are more forgiving and more flattering to faces, which is why they dominate portrait-style shorts.

When generating footage, describe the optical result rather than the lens name: background compressed close behind the subject, soft falloff, minimal distortion or wide environmental view, subject small in frame, deep space behind. Physical descriptions generate more predictable results than brand references, and they survive a model swap. The AI video generator responds to that kind of language consistently.

Grade and texture in one pass

Grading is the last ten percent of the work that reads as the final twenty. A simple, repeatable pass beats a complex one applied inconsistently.

Four steps cover most short-form needs:

  1. Normalize. Match exposure and white balance across shots before doing anything creative. This is the step people skip, and it is the reason their grade looks uneven.
  2. Set contrast. Add a gentle S-curve. Protect the highlights, let the shadows fall slightly.
  3. Tint the extremes. Push shadows toward one hue and highlights toward its complement. Teal and orange is famous, but the underlying principle is cool shadows with warm skin, and it generalizes to any palette.
  4. Add texture. A light grain or halation layer helps generated footage blend with real footage and disguises the overly clean edges that give synthetic output away.

Do all four in a single session. Judging a final image is easier than judging an intermediate one, and grading shot by shot is how palettes drift.

A workflow you can repeat

Cinematic results come from process, not inspiration. Here is a sequence that works whether you shoot, generate, or combine both.

Step 1: write the one-page look brief

Four lines: palette, light direction, lens family, movement rule. Everything downstream references this page. If you change it mid-project, change it for the whole piece rather than one shot.

Step 2: develop the look on stills

Stills are cheap and fast. Produce ten frames that represent your intended look and compare them side by side. If two frames look like different films, your brief is not specific enough. An AI image generator is useful here because you can iterate on composition and lighting without spending time on motion.

Step 3: build a shot list with intention columns

A shot list with only descriptions is a shopping list. Add two columns: what the shot does dramatically, and which of the four look rules it uses. Any shot that cannot fill those columns gets cut. This is the single most effective filter for first drafts.

Step 4: batch by setup, not by story order

Group shots by lighting setup and by movement type. Shooting every push-in together and every static frame together saves a surprising amount of time and keeps visual logic consistent, because you are not re-lighting the same idea three times in a day. The same applies to generation: batch prompts that share a look description so the output stays coherent across a sequence.

Step 5: edit for rhythm, not completeness

Cinematic pacing in vertical video is faster than film and slower than a montage. Cut on movement, cut on sound, and cut before the viewer has fully read the frame. Most first edits run twenty percent too long, and trimming that excess is usually the difference between a clip people finish and a clip people scroll past.

Step 6: finish in one grading session

Color and texture belong in the same pass, judged together on the final timeline. Splitting them creates intermediate versions that look right alone and wrong in sequence.

Shooting, generating, or both: a decision checklist

Both approaches reward the same cinematographic knowledge, but they have different strengths. Use this as a decision checklist rather than a rulebook.

  • Choose live shooting when performance, physical interaction, or a specific real location is the point. Hand nuance, eye micro-movements, and breath still read differently on camera.
  • Choose generation when the shot is impossible, expensive, geographically distant, or when you need many variations of one look quickly. It also wins for abstract transitions, stylized establishing shots, and concept frames you will later match on set.
  • Combine them when you need coverage. Generate the wide establishing shot, shoot the close performance, then grade both with a shared palette and a shared grain pass.

Whichever route you take, keep the look brief identical. The most common failure in hybrid projects is that generated shots and real shots obey different rules, and audiences feel the seam even when they cannot name it. Starting from a video template can help you inherit a coherent structure instead of inventing one per shot.

Mistakes that flatten a frame and how to fix them

  • Everything is in focus. Increase subject-to-background distance, open the aperture, or describe shallow depth explicitly in the prompt.
  • Flat frontal light. Move the key forty-five degrees off-axis and add negative fill on the shadow side.
  • The camera moves for no reason. Name the emotional job of each move. If there is no job, lock the frame.
  • Composition ignores the top third. Place eyelines on the upper third line and use the top of the frame for intentional texture or space.
  • Palette drifts between shots. Apply tint rules across the whole sequence rather than grading individual clips.
  • Generated shots look pasted together. Reuse identical look language across prompts and add grain plus slight lens imperfection in post.
  • Too much motion in a small frame. Halve the move distance and let the subject carry the movement instead.

FAQ

How many shots should a cinematic vertical short contain?

For a thirty to sixty second piece, eight to fifteen shots is a comfortable range. Fewer shots mean each one must be composed and lit exceptionally well. More shots mean the viewer never settles, which can work for high-energy edits but rarely feels cinematic.

Do I need a gimbal for smooth movement?

Not necessarily. A slow push on a stable surface, a slider improvised from a shelf edge, or a locked-off shot with a moving subject will often look better than a shaky attempt at a complex move. Smoothness matters less than motivation.

Can generated footage look genuinely cinematic?

Yes, if you feed it cinematographic structure: named light direction, explicit depth layers, described optical behavior, and a consistent palette. Prompts that only describe content produce generic images. Prompts that describe light and optics produce footage you can cut against real material.

What is the single biggest upgrade for a beginner?

Put the key light off-axis and darken everything else. Directional light alone will improve more shots than a new lens or a new camera body.

How do I keep a series visually consistent?

Keep a written look brief and reuse it verbatim across episodes. Consistency across a series builds recognition faster than any individual impressive shot, because it teaches the audience what your work looks like before they know your name.

Should I crop horizontal footage to vertical?

Only for shots composed with a central subject and generous vertical space. Otherwise, reshoot or regenerate. Cropping 16:9 to 9:16 usually destroys the composition that made the original shot work, and the result reads as a compromise rather than a choice.

How long should a single shot hold before cutting?

Long enough for the viewer to read the frame and short enough that they never get comfortable. In vertical edits, that usually lands between one and three seconds for information shots and up to six seconds for a hero frame with movement inside it.

Put the look brief into motion

Cinematography is a small number of decisions applied relentlessly. Compose for the tall frame, layer your depth, move with purpose, light from one direction, and control what stays sharp. That is the entire craft, repeated until it becomes instinct.

When you are ready to put those decisions into motion, take your look brief into Orelon and start generating shots that follow it. Begin with a single hero frame, compare your options across models through the alternatives pages, and keep the Orelon blog open for the next technique you want to master. The frame is yours to direct.