Guides / Prompting

Camera moves: what each one does and the words that get it

A camera move decides what the viewer looks at and how they feel about it. Here are twelve, each with a clip made in NOLGIA, the exact prompt that got it, and the move it is most often mistaken for.

  • 25 min read
  • Updated

AI video looks cheap when the camera does nothing, or does something nobody chose. A camera move fixes that, but only if the model knows which move you mean, and most of us use the words loosely: "zoom in" for a dolly, "pan up" for a tilt, "spin around" for an orbit.

Here are the twelve moves people ask for most, from the dolly in to bullet time. For each: what the camera does, how it feels, a clip made in NOLGIA with the prompt that got it, the words to write, and the move it gets confused with. Then the other words worth knowing, the composer's move picker, and six rules for writing any move.

The twelve moves at a glance

A montage of the twelve camera moves, a second and a half of each: an office, a driveway, a coast road, a hotel lobby, a night street, a bar, a birthday dinner, a shopping street, a crowd of photographers, a red sports car, a hotel corridor and a pool hall
All twelve moves, made with Seedance 2.5 in NOLGIA. A second and a half of each, in the order of this guide, without sound.

Every move clip here was made the same way: a start frame made in NOLGIA with Nano Banana Pro (GPT Image 2.5 Sunburst for the pan), then Seedance 2.5 image to video at 720p, 9:16 and 5 seconds, without sound. Each section quotes both prompts exactly as sent, and shows the takes that missed.

MoveWhat the camera doesUse it to
Dolly inTravels toward the subjectPull the viewer into a thought or a decision
Dolly outTravels away from the subjectReveal where they are, or leave them alone
PanTurns left or right from one spotFollow something across a place
TiltTips up or down from one spotShow height, or size someone up from shoes to face
Whip panA pan so fast the frame blursJump to something new with energy, or hide a cut
Crash zoom inThe lens snaps tightLand a reaction, a detail or a punchline
Crash zoom outThe lens snaps wideReveal the room all at once, for a laugh or a shock
Tracking shotTravels with a moving subjectStay with someone on the move
HandheldHeld by a person, with natural swayMake it feel real and unplanned
OrbitCircles the subjectMake a moment feel big
Dolly zoomTravels one way while the lens zooms the otherShow the ground shifting under someone
Bullet timeTime stops while the camera keeps movingFreeze a peak moment and look around it

Tip:Make the start frame for the move. The first frame decides what the move can show: a tilt up needs a frame that starts low, a crash zoom out a frame that hides the room. Describe people plainly as ordinary people; one of our first frames came back with someone who looked like a real celebrity, and the plainer description in the whip pan frame fixed it. If a frame comes back as two pictures stacked, ask for one continuous photograph, as the dolly zoom frame does.

Dolly in

What the camera does. The whole camera travels forward toward the subject, as if it were on a track. The frame tightens, and things near the lens slide past faster than the background, so you feel yourself moving into the room.

How it feels. It pulls the viewer closer, the way you lean in when something matters. Use it on a decision, a realization, or the line that counts.

A wide shot of a Manhattan office at golden hour: the camera travels from the doorway toward a man in a grey suit counting money at his desk, a woman in a black dress at the window behind him
Dolly in, made with Seedance 2.5. The camera travels from the doorway toward his desk while he keeps counting; the woman at the window grows in the frame with him.
The prompt that got it (Seedance 2.5, 5 seconds)
Dolly in. The camera glides slowly and steadily straight forward on a dolly, from this wide shot of the office toward the man at the desk, ending on a medium close-up of him counting the money. A real camera move, not a zoom: the perspective changes as the camera gets closer. He keeps counting; the woman stays at the window.
Its start frame (Nano Banana Pro)
1990s film still on 35mm: a Manhattan executive office at golden hour. A man in a grey double-breasted suit sits at a large mahogany desk counting a stack of hundred-dollar bills; a woman in a black dress stands by the floor-to-ceiling window behind him, the skyline outside. Wide shot from the doorway, symmetrical, warm window light, true-to-life colour, fine film grain. No text.

Write it like this:

  • The camera slowly dollies in toward him, from a medium shot to a close-up on his face.
  • Slow push in on her as she reads the letter.

Watch for this:A dolly in is not a zoom in. A zoom changes the lens while the camera stays put: everything grows evenly and the depth flattens. A dolly moves the camera, so the space opens around the lens. Write "dolly in" or "push in" and say the camera moves forward, as the prompt above does: "A real camera move, not a zoom".

Dolly out

What the camera does. The camera travels backward, away from the subject, and the frame widens to show more of the place around them.

How it feels. A reveal: where they are, who else is there, or how alone they are. A slow dolly out at the end of a scene says goodbye.

A man in a cream linen suit and sunglasses in a sunny mansion driveway: the camera starts close on his face and backs away to show a black sedan, a stone fountain and the house
Dolly out, made with Seedance 2.5. The camera starts close on his face and backs away, bringing in the black sedan, the fountain and the house.
The prompt that got it (Seedance 2.5, 5 seconds)
Dolly out. The camera starts close on his face and moves smoothly straight backward on a dolly, revealing the black sedan, the driveway and the fountain around him. A real camera move, not a zoom. He stays still and looks just past the lens. Slow and steady.
Its start frame (Nano Banana Pro)
1990s film still on 35mm: close-up of a man in a cream linen suit and gold-rimmed sunglasses, standing in a mansion driveway in bright Los Angeles afternoon sun; a black 1990 sedan and a stone fountain softly visible behind him. True-to-life colour, fine film grain. No text.

Write it like this:

  • The camera slowly dollies out from a close-up of her face to a wide shot of the empty street.
  • Pull back from his hands to reveal the whole workshop.

Watch for this:Say what it reveals. Dolly out, pull back and pull out are the same move, and it is only as good as what it uncovers. Name it, as the prompt names the sedan, the driveway and the fountain, or the model widens onto whatever it invents.

Pan

What the camera does. The camera stays in one spot and turns left or right, like turning your head.

How it feels. The viewer looks around a space from one point of view, or follows someone across it. A slow pan builds a place; a pan that follows a moving car or a walker keeps them in the frame.

A coast road at sunset above the ocean: from one spot by the road, the camera turns right to follow a small red convertible, the cliffs sliding out of the frame
Pan right, made with Seedance 2.5. From one spot by the road, the camera turns with the red convertible; the cliffs slide out of the frame and the open sea takes over.
The prompt that got it (Seedance 2.5, 5 seconds)
Pan right. The camera stays in one spot at the roadside and only turns horizontally to the right, following the red convertible as it drives from left to right along the coast road, so the cliffs and the ocean sweep past behind it. The camera does not travel or move closer. No camera equipment or people appear at the roadside.
Its start frame (GPT Image 2.5 Sunburst)
An upright, correctly oriented vertical photograph, 1990s film still on 35mm, golden hour on the Pacific Coast Highway. The sky with a low sun is at the top, the ocean and rocky cliffs in the middle, and in the lower third the two-lane coast road runs horizontally across the frame from left to right. A small red convertible is at the far left edge of the road, driving toward the right. Shot from a tripod at the roadside. True-to-life colour, fine film grain. No text.

The pan took three takes. The first used a different start frame, a diner with a waitress walking down the aisle toward the lens; the second used this coast road but said the camera stood "fixed on a tripod at the roadside".

Take 1, a diner with a black and white floor: a waitress in a pink uniform walks down the aisle toward the camera carrying plates, and the camera hardly turns
Take 1. She walks toward the lens, so there is nothing for a pan to follow: the camera hardly turned, and it reads as her approaching.
Take 2, the coast road at sunset: the camera turns with the red convertible, and a camera tripod appears standing by the road
Take 2. The pan worked, but the prompt put the camera "on a tripod at the roadside", and a tripod appeared there.

Tip:Give a pan something that crosses the frame. A pan follows movement from side to side. Action that comes toward the camera gives it nothing to turn with. And describe what the camera does, not the gear it stands on: the kept take says the camera "stays in one spot" and adds "No camera equipment or people appear at the roadside." It used the same start frame as take 2; only the video prompt changed.

Write it like this:

  • From a fixed position, the camera pans slowly from left to right across the diner, from the counter to the window.
  • The camera pans to follow her as she crosses the room.

Watch for this:A pan is not a truck. A pan turns the camera. A truck slides the whole camera sideways without turning, so things near the lens slide past faster than the background. If the camera should travel along a row of shops, write "the camera trucks left", not "pans left".

Tilt

What the camera does. The camera stays in one spot and tips up or down, like nodding.

How it feels. Tilting up from the shoes to the face sizes someone up and makes them look important. Tilting up a building shows its height; tilting down brings the viewer back to the ground.

A hotel lobby with a marble floor and a brass revolving door: the camera tips up from a man's polished black shoes, past his navy suit, to his face as he straightens his tie
Tilt up, made with Seedance 2.5. The camera tips up from his shoes, past the suit, to his face as he straightens his tie.
The prompt that got it (Seedance 2.5, 5 seconds)
Tilt up. The camera stays in one place and tilts slowly upward, from his polished shoes, up his suit trousers and jacket, to his face as he straightens his tie. No sideways or forward movement, only the vertical tilt.
Its start frame (Nano Banana Pro)
1990s film still on 35mm: low angle close on a man's polished black oxford shoes and dark suit trousers standing on a marble hotel lobby floor, a brass revolving door and crystal chandeliers softly behind, warm morning light. Only his legs and shoes are in frame. True-to-life colour, fine film grain. No text.

Write it like this:

  • From a fixed position low to the floor, the camera tilts up from his polished shoes to his face.
  • Slow tilt down from the treetops to the cabin.

Watch for this:There is no pan up. Up and down is a tilt. And if the camera should rise or sink without tipping, that is a pedestal, or a crane, jib or boom when it rises high on an arm. Write "tilts up" for tipping and "rises" or "cranes up" for climbing.

Whip pan

What the camera does. A pan so fast that the picture smears into motion blur for a moment, then snaps to rest on something new.

How it feels. Energy and surprise. It yanks the viewer's attention from one thing to another, and editors use the blur to hide a cut between two shots.

A city sidewalk at night outside a nightclub with a doorman: the camera whips into a blur of streetlights and stops on a young man in a black leather jacket walking toward it
Whip pan, made with Seedance 2.5. The camera holds on the street for a beat, whips away in a blur of streetlights, and stops sharply on the young man in the leather jacket.
The prompt that got it (Seedance 2.5, 5 seconds)
Whip pan. The camera holds on the doorman for a moment, then swings very fast to the right with strong motion blur and stops sharply on the young man in the leather jacket walking toward the camera. One quick horizontal whip from left to right.
Its start frame (Nano Banana Pro)
1990s film still on 35mm, at night: a Sunset Boulevard sidewalk outside a nightclub under warm tungsten streetlights and a brass entrance awning. A doorman in a black overcoat stands on the left; on the right a lean man in his late twenties with short curly dark hair, a thin moustache and a black leather jacket walks toward the camera. An ordinary, unfamous-looking person, not resembling any celebrity. Warm amber light only, no neon. True-to-life colour, fine film grain. No text.

Write it like this:

  • He walks toward the camera; the camera whip pans to the right, the frame blurring, and lands on three friends laughing by the door.
  • A fast whip pan from her face to the window, with motion blur.

Watch for this:Say where it lands. Half of a whip pan is the landing. Name the second subject and where it is, or the model blurs into nothing in particular. And say "whip pan" or "very fast": a plain pan comes back smooth.

Crash zoom in

What the camera does. The lens zooms in very fast, from a wide or medium frame to a tight one, while the camera stays where it is.

How it feels. A jolt. It points at something the viewer must not miss: a reaction, a detail, a punchline. Comedy and old action films use it a lot.

A woman in a black slip dress leaning on a hotel bar, chin on her hand: the frame rushes in from the bar to an extreme close-up of her eyes
Crash zoom in, made with Seedance 2.5. The lens rushes from the bar to her eyes. It took about two seconds: fast, but slower than the half second the prompt asked for.
The prompt that got it (Seedance 2.5, 5 seconds)
Crash zoom in. The camera does not move; the lens zooms in very fast, in under half a second, from this medium wide shot to an extreme close-up of her eyes as she smiles. A sudden, punchy zoom, then hold.
Its start frame (Nano Banana Pro)
1990s film still on 35mm: a woman in a black slip dress leans on a polished wooden bar in a hotel bar, chin resting on her hand, looking toward the camera; brass lamps and bottles behind her, warm light. Medium wide shot. True-to-life colour, fine film grain. No text.

Write it like this:

  • Sudden crash zoom in on her face as she hears the news.
  • The camera holds still, then snaps into a fast zoom on the gold watch on the bar.

Watch for this:A crash zoom is not a fast dolly in. A dolly travels, so the space opens up around the lens. A crash zoom is the lens alone, and everything flattens. Say "the camera does not move", as the prompt does, and ask for the speed you want; models tend to land a crash zoom in a second or two, so if it must snap, say it happens in the first second and then holds.

Crash zoom out

What the camera does. The reverse: the lens snaps from a tight frame to a wide one, showing everything around the subject at once.

How it feels. A reveal played for a laugh or a shock: the stern face that turns out to be at a birthday party, the hero who turns out to be surrounded.

A pink birthday cake with lit candles in a wood-panelled steakhouse: the frame snaps back from the cake to the whole long table, a man in a dark suit at the end and his friends applauding
Crash zoom out, made with Seedance 2.5. From the cake to the whole table in about a second and a half, his friends applauding on both sides.
The prompt that got it (Seedance 2.5, 5 seconds)
Crash zoom out. The camera does not move; the lens zooms out very fast, in under half a second, from the close-up of the birthday cake to a wide shot of the whole dinner table, the man and his friends applauding. A sudden zoom, then hold.
Its start frame (Nano Banana Pro)
1990s film still on 35mm: close-up of a pink frosted birthday cake with lit candles on a white tablecloth in a wood-panelled steakhouse; a man in a dark suit leans in behind the cake, slightly out of focus, warm lamp light. True-to-life colour, fine film grain. No text.

Write it like this:

  • Start on an extreme close-up of his eyes, then a sudden crash zoom out to reveal the whole restaurant singing to him.
  • Fast zoom out from her face to a wide shot of the packed stadium.

Watch for this:Start tight on purpose. The reveal needs a first frame that hides it, which is why this one starts on the cake. Name what the wide shot shows: the bigger the gap between the two, the harder it lands. The composer's move row has no zoom out, so write this one.

Tracking shot

What the camera does. The camera travels with a moving subject, ahead of them, behind them or beside them, and keeps them about the same size in the frame.

How it feels. The viewer goes along. It is the move for walking and talking, chases and entrances.

A woman in a grey suit with a black handbag walks along a sunny shopping street: the camera moves sideways beside her at her pace while the shop windows and palm trees slide past behind her
Tracking shot, made with Seedance 2.5. The camera keeps pace beside her, holding her the same size, while the shop windows slide past behind.
The prompt that got it (Seedance 2.5, 5 seconds)
Tracking shot. The camera moves sideways alongside her at walking pace, keeping her the same size in the same place in the frame, while the shop windows and palm trees slide past behind her. Smooth, like a camera on a dolly track.
Its start frame (Nano Banana Pro)
1990s film still on 35mm: a woman in a grey power suit with a black handbag walks along a Beverly Hills shopping street past boutique windows in bright morning sun, seen from the side at walking height, medium shot. True-to-life colour, fine film grain. No text.

Write it like this:

  • Tracking shot: the camera moves backward ahead of him as he walks toward it down the corridor, holding him from the knees up.
  • The camera follows behind her through the crowd at walking pace.

Watch for this:A tracking shot travels; a pan only turns. A pan can follow a walker too, but from one spot, so they get smaller or bigger as they go. A tracking shot moves with them. Say which side the camera is on, keep the distance ("keeping her the same size") and match the pace ("at walking pace").

Handheld

What the camera does. A person holds the camera, so it sways, breathes and corrects itself a little. Gentle handheld drifts; rough handheld shakes and lurches with the action.

How it feels. Real, close and unplanned, like footage someone caught rather than staged. Use it for documentary, behind the scenes, arguments, and anything that should feel like it really happened.

Night outside a nightclub, seen from inside a crowd of photographers with flash cameras: the frame shakes and reframes as a man in a black leather jacket steps out of a black car and pushes through
Handheld, made with Seedance 2.5. Shot from inside the crowd: the frame shakes and reframes as he steps out of the car and pushes through the photographers.
The prompt that got it (Seedance 2.5, 5 seconds)
Handheld camera. The camera is held by a photographer in the crowd: small natural shakes, slight drifts and quick reframes as the man steps out of the car and pushes through the flashing cameras. Documentary feel, not smooth.
Its start frame (Nano Banana Pro)
1990s film still on 35mm, at night: paparazzi with flash cameras crowd a nightclub entrance as a man in a black leather jacket steps out of a black town car; warm streetlights and camera flashes. Seen from inside the crowd at shoulder height. True-to-life colour, fine film grain. No neon. No text.

That is the rough kind. Here is the gentle kind, also made with Seedance 2.5 in NOLGIA: two people at a table by a pool, talking to whoever holds the camera.

A man and a woman in swimwear sit at a glass patio table under a covered terrace, a pool and palm trees behind them; the camera starts wide, drifts toward each of them as they talk, then holds as they turn to each other
Realistic handheld, made with Seedance 2.5. 14 seconds, 9:16, with its own sound. The camera starts wide, drifts toward her as she talks, eases back for him, then keeps rolling as they turn to each other.
The prompt that got it (Seedance 2.5, 14 seconds, 9:16, 720p)
SEEDANCE 2.5 - REALISTIC HANDHELD POOL CONVERSATION

Use the provided character references and location reference exactly.

Create a highly realistic early-1990s American backyard pool scene.

The couple is seated around the outdoor glass-top patio table shown in the location reference, with the swimming pool clearly visible behind them.

COMPOSITION:

Start with a WIDE HANDHELD SHOT that clearly establishes the entire setting.

Show the couple's seated bodies, the glass-top patio table, surrounding patio chairs, flower centerpiece, covered patio and columns, swimming pool, backyard landscaping, and house.

The pool must remain clearly visible in the background.

The MAN is seated somewhat sideways, initially with his body oriented away from the camera. As the shot begins, he naturally turns his head and upper body back toward the camera.

The WOMAN is also seated slightly sideways. She naturally turns her face toward the camera.

They should NOT look like they have been positioned for a formal interview.

CAMERA:

The camera is operated by a real person standing several feet away.

Use authentic handheld camera behavior:
- subtle natural hand shake
- tiny framing corrections
- slight drifting
- small imperfect movements
- occasional gentle reframing

Do NOT perfectly stabilize the footage.

Begin relatively wide so the pool, table, chairs, and environment are clearly visible.

As each person speaks, the camera can naturally move slightly closer toward them, transitioning from wide to medium framing.

When the woman speaks, subtly move toward her.

When the man speaks, naturally reframe toward him.

Do not make these dramatic cinematic zooms. They should feel like the camera operator is simply adjusting their position while filming.

DIALOGUE:

For the ENTIRE SCRIPTED PORTION, both people are speaking DIRECTLY TO THE CAMERA.

WOMAN:
“Oh, and don’t forget speed.”

MAN:
“Real-time for natural movement.”

WOMAN:
“Slow motion when you want the moment to actually hit.”

MAN:
“Tell it where the camera moves.”

WOMAN:
“Control the camera…”

MAN:
“and you control what people look at.”

PERFORMANCE:

They are talking naturally to the person holding the camera.

Natural conversational timing.

Natural pauses and breathing.

Natural blinking.

Subtle facial expressions.

Small hand gestures.

Natural posture shifts.

They should occasionally glance briefly at each other, but their attention remains primarily on the CAMERA throughout the scripted dialogue.

Do NOT make them sound like presenters reading a script.

Do NOT make them overly enthusiastic.

Do NOT make them constantly smile.

The delivery should feel relaxed, spontaneous, confident, and conversational.

The second person's line should naturally follow the first person's thought rather than feeling like they are waiting for a cue.

REALISTIC HUMAN DETAIL:

Extremely realistic skin texture with visible natural pores and subtle imperfections.

Natural facial asymmetry.

Realistic eyes and blinking.

Natural mouth movement and accurate lip synchronization.

Realistic teeth and tongue movement while speaking.

Natural hair movement.

Realistic clothing texture and wrinkles.

Natural breathing and subtle body movement.

No plastic skin.
No beauty-filter appearance.
No perfect artificial faces.
No AI-looking eyes.

LIGHTING:

Use the natural outdoor afternoon lighting from the location reference.

Realistic sunlight.
Natural shadows.
Natural skin tones.
Believable exposure.

NO cinematic lighting.
NO artificial rim lighting.
NO dramatic color grading.
NO glossy commercial lighting.

IMPORTANT ENDING:

Immediately AFTER the man finishes the final scripted line:

“and you control what people look at.”

There is a short natural pause.

Then they STOP talking to the camera.

They naturally turn toward EACH OTHER.

The woman turns toward the man with a genuine happy smile.

The man turns toward the woman and smiles back.

They begin having a casual, happy conversation BETWEEN THEMSELVES.

They are no longer addressing the camera.

They should look completely unaware of the camera for this final portion, as if the scripted recording is over and they have simply gone back to talking to each other.

Their conversation does NOT need to be intelligible. Natural speech-like mouth movement is enough.

They can laugh softly, smile, make small gestures, react to each other, and continue chatting naturally.

The camera operator keeps recording them from the same handheld position.

Do NOT have either person look back at the camera during this final interaction.

Keep enough distance to show their bodies, the table, chairs, and pool while they talk to each other.

The transition from scripted dialogue to their private conversation must feel completely natural - no dramatic pause, no posed smile, no obvious "outro" behavior.

The final footage should feel like the camera accidentally kept rolling after they finished talking to the viewer.

NO CUTS.
NO SUBTITLES.
NO TEXT.
NO BACKGROUND MUSIC.
NO EXTRA SCRIPTED DIALOGUE.

The final interaction can be covered by a logo, so intelligible audio is not required after the scripted portion. Camera: handheld with visible natural sway and micro-shake, the operator adjusting the framing as the action unfolds, a documentary feel.

The only change to that prompt is that long dashes are typed as hyphens. It never says "shaky camera". It describes a real operator instead: "subtle natural hand shake", "tiny framing corrections", "slight drifting", and it sets a limit: "Do not make these dramatic cinematic zooms." Its last sentence is the Handheld camera move at Medium, word for word: the sentence the camera move row adds when you pick it.

Write it like this:

  • Handheld camera with natural sway and small framing corrections, a documentary feel.
  • Shaky handheld, the camera jostled by the crowd around him.

Watch for this:Handheld is not a Steadicam or a gimbal. Those are carried by a person too, but they float smoothly. For a moving camera without shake, write "smooth gimbal shot" or "Steadicam". For shake, say how much.

Orbit

What the camera does. The camera travels in a circle, or part of one, around the subject, keeping them in the middle of the frame while the background wheels past behind them.

How it feels. Importance. It turns someone standing still into an event, and shows them from more than one side. A slow orbit feels like admiration; a fast one, like a spin of the head.

A woman in a white skirt suit and sunglasses stands beside a red sports car in a palm-lined driveway: the camera swings around her, the garage behind her giving way to a hedge and palm trees
Orbit, made with Seedance 2.5. The camera swings around her while she stays in the middle of the frame; the garage gives way to the hedge and the palms. It travelled less than the half circle the prompt asked for.
The prompt that got it (Seedance 2.5, 5 seconds)
Orbit. The camera circles around her in a smooth half circle from left to right at the same distance, keeping her in the centre of the frame, so the red car and the palm trees swing past behind her. She turns her head slightly to follow the camera.
Its start frame (Nano Banana Pro)
1990s film still on 35mm: a woman in a white tailored skirt suit with gold buttons stands beside a red 1990s Italian sports car in a sunny palm-lined Beverly Hills driveway. Full-length shot, subject centred. True-to-life colour, fine film grain. No text.

Write it like this:

  • The camera slowly orbits around him, a half circle from his left to his right, keeping him centered as the harbour comes into view behind him.
  • A quarter orbit to the right around the car, at the same height.

Watch for this:An orbit is not a pan. In a pan the camera stays put and turns, so the subject slides across the frame. In an orbit the camera travels around the subject, so the subject stays centered and the background moves. Say "circles around", the direction, and how far: a quarter, a half or a full circle. A shallow orbit is also called an arc.

Dolly zoom

What the camera does. The camera travels one way while the lens zooms the other way, at a speed that keeps the subject the same size. The subject holds still in the frame while the background stretches away or closes in behind them. It is also called the vertigo effect.

How it feels. The ground shifting under someone: dread, shock, a sudden realization. It unsettles because nothing in life looks like that.

A man in a navy suit stands in a wood-panelled hotel corridor staring at the camera: he stays about the same size while the corridor behind him stretches longer and the far door falls away
Dolly zoom, made with Seedance 2.5. He stays about the same size while the corridor behind him stretches longer and the far door falls away.
The prompt that got it (Seedance 2.5, 5 seconds)
Dolly zoom, the Vertigo effect. The camera moves backward along the corridor while the lens zooms in at the same time, so the man stays exactly the same size in the frame while the corridor behind him stretches and the far door seems to rush away. Slow and unsettling. He stays still and stares at the camera.
Its start frame (Nano Banana Pro)
One single continuous photograph, not a collage or split image. 1990s film still on 35mm, vertical frame: a man in his forties with greying hair, wearing a dark navy suit, stands in the middle distance of a long hotel corridor, seen from waist up to above his head, centred. Behind him the corridor recedes in symmetrical one-point perspective: wood-panelled walls, brass wall sconces on both sides, patterned carpet, a closed door at the far end. Warm light, true-to-life colour, fine film grain. No text.

Write it like this:

  • Dolly zoom: the camera pushes in while the lens zooms out, so she stays the same size and the street behind her stretches away.
  • Vertigo effect on his face: the camera pulls back as the lens zooms in, and the room closes in behind him.

Watch for this:Describe the result, not only the mechanics. Models often read "dolly zoom" as a plain push in, so spell it out: the camera moves one way, the lens zooms the other, the subject stays the same size, and say what the background does. Strictly, a background that stretches away comes from moving in while zooming out; the prompt above has the directions the other way round, but it also names the result it wants, "the corridor behind him stretches", and that is what came back.

Bullet time

What the camera does. Time stops, or nearly stops, while the camera keeps moving, around or through the frozen moment. The action hangs in the air: water, glass, cards, a jump.

How it feels. The viewer gets to walk around a peak moment. It turns half a second into a set piece.

Take 1: "time almost freezes"

Take 1 in the pool hall: the woman throws the poker chips and they fly out of her hand as in an ordinary throw, the camera barely moving

Take 2, kept: "time stops completely"

Take 2, kept: the poker chips hang still in the air in front of her while the camera moves in and around her
Bullet time, made with Seedance 2.5. The first prompt said time "almost freezes" and then "resumes", and the chips simply flew out of her hand. The second said time stops completely and nothing moves except the camera: the chips hang still while the camera moves in and around her.
The prompt that got it (Seedance 2.5, 5 seconds)
Bullet time. Time stops completely: she and the poker chips are frozen in mid-air, perfectly still, while the camera alone travels smoothly in a quarter circle around her from left to right, showing the frozen chips from a new angle. Nothing in the scene moves except the camera.
Its start frame (Nano Banana Pro)
1990s film still on 35mm: in a pool hall, a woman in a black leather jacket flings a handful of red, white and blue poker chips toward the camera; the chips are mid-air, warm lamp light over the green felt table. True-to-life colour, fine film grain. No neon. No text.

Write it like this:

  • Bullet time. Time stops completely as the glass shatters; the pieces hang in the air while the camera slowly circles the table. Nothing moves except the camera.
  • Time stops completely at the top of his jump; the camera keeps moving around him at normal speed.

Watch for this:Bullet time is not slow motion. In slow motion everything slows, the camera included. In bullet time the action stops and the camera does not, so write two speeds: the action frozen, the camera moving at its own pace along a path you name. The composer's move row has no bullet time, so write this one.

More moves, shot sizes and angles

The moves above cover most shots. These are the other words worth knowing: more ways for the camera to travel, then how close it is and where it stands.

Move or termWhat the camera doesWrite it like this
Truck (or crab)Slides sideways, parallel to the scene, without turningthe camera trucks left along the row of shops
PedestalRises or sinks straight up or down, without tippingthe camera rises slowly from her hands to her face
Crane, jib or boomSwings up, down or over the scene on an arm, often highcrane up from the street to a high view of the square
ArcA partial orbit: a curve around the subjecta slow arc to the right around him
Push in, pull outOther names for dolly in and dolly outslow push in on her face
Steadicam or gimbalCarried by a person but floating smoothly, with no shakesmooth gimbal shot following him up the stairs
Rack focusThe camera holds still and the focus moves from one thing to anotherrack focus from the ring in the foreground to her face
FPV or droneFlies freely, often fast, banking around thingsFPV drone shot flying low over the river and up over the bridge
Top down or overheadLooks straight down from abovetop down shot looking straight down at the table
Locked off or staticDoes not move at all; only the scene movesthe camera does not move at all; only she moves
Slow motionThe action plays slower than life; the camera may move or notin slow motion, the water hanging in the air
Other moves and camera terms.
Shot or angleWhat it showsWrite it like this
Extreme wideThe whole place, people small in itextreme wide shot of the valley, the hiker tiny on the ridge
WideThe whole person and the space around themwide shot, full body, the whole kitchen in frame
MediumFrom the waist upmedium shot from the waist up
Close-upThe face, or one object filling the frameclose-up on his face
Extreme close-upOne detail: eyes, lips, a hand, a keyholeextreme close-up on her eyes
Low angleFrom below, looking up: makes people look biggerlow angle looking up at him
High angleFrom above, looking down: makes people look smallerhigh angle looking down at her at the desk
Dutch angleThe camera rolled sideways so the horizon slants: off balanceDutch angle, the horizon tilted
Over the shoulderPast one person's shoulder at another, for conversationsover the shoulder shot from behind him, looking at her
POVWhat a character sees, through their eyesPOV shot through his eyes as he opens the door
Shot sizes and angles.

Tip:A full direction is three choices. How close, where the camera stands, and how it moves: "Low angle, wide shot, slow push in toward him" says all three, and leaves the model nothing to pick at random.

Pick the move or write it

In the video composer, under the prompt, is the Camera move row: twenty moves, each with a strength: Subtle, Medium, and Strong. Pick one, and the row shows the exact sentence it will add, after "Added to your prompt when you generate". Picking never edits what you typed: NOLGIA adds that sentence after your prompt when you press Generate video. It works on every video model, for text or a start image, one shot or several.

MovePick in the Camera move rowOpen the composer with it picked
Dolly inPush-inPush-in
Dolly outPull-backPull-back
PanPan left or Pan rightPan right
TiltTilt up or Tilt downTilt up
Whip panPan left or Pan right at StrongPan right, strong
Crash zoom inZoom in at StrongZoom in, strong
Crash zoom outNot in the rowWrite it
Tracking shotTracking shotTracking shot
HandheldHandheldHandheld
OrbitOrbit left or Orbit rightOrbit right
Dolly zoomDolly zoomDolly zoom
Bullet timeNot in the rowWrite it
Each link opens the video composer with that move and strength already picked.

The row also has moves from the table above: Truck left and Truck right, Crane up and Crane down, Rack focus, Parallax drift, Drone fly-through and Locked off.

Pick it when you want one standard move and do not mind the wording. Write it when the move needs a start and an end ("from his shoes to his face"), a moment ("after she sits down"), a landing ("stops sharply on the young man"), or when the row does not have it, like a crash zoom out or bullet time. The twelve move clips in this guide were written, not picked. Do one or the other: if you write your own camera direction, leave the row empty, or press No camera move, so two directions do not argue.

Enhance, under the prompt, rewrites your prompt for the model you picked and plans the shot around the move you picked in the row. The price is on the button, and Revert puts your words back.

Tip:The strength is in the words. Subtle, Medium and Strong change the sentence that gets added, not the price. At Strong, Pan right becomes a whip pan ("a fast whip pan to the right from a fixed position, the frame blurring with motion before snapping to rest on the new view") and Zoom in becomes a crash zoom.

When we tested the row in September, eight moves from one still on Wan 3.0, six came back as picked. Dolly zoom came back as a plain push in at Medium and at Strong, Pan right came back as an orbit, and Pull-back only pulled back at Strong. Single runs, not a measurement, but the lesson holds: a move is a request, so play the clip, and when it misses, run it again or try the next strength up.

A move adds nothing to the price: you pay the model's rate for the clip, shown on the Generate video button before anything runs. The live published rate for Seedance 2.5, the model behind every clip in this guide:

  • Seedance 2.5Pro and up

    Per 5s clip

    480p 26 credits · 720p 56 credits · 1080p 137 credits

    1080p renders at 16:9 and 9:16

Read live from the catalog when this page loads. Every price is shown before you generate. See every rate.

Six rules for writing a move

  1. Name the move, then describe it. Every move prompt in this guide opens with the move's name ("Dolly in.", "Whip pan.") and then says what the camera physically does, because the name alone is often read loosely.
  2. One move per shot. "Dolly in, then orbit, then crane up" in five seconds comes back as a wobble that is none of them. If you need two, give the second its own moment ("after he sits down"), and consider two shots instead.
  3. Say where the camera starts and where it ends, and how fast it gets there: "from his polished shoes, up his suit trousers and jacket, to his face", "Slow and steady". Without an end, the model guesses.
  4. Name what stays still. "He keeps counting; the woman stays at the window." "He stays still and looks just past the lens." When the people hold still, the camera is clearly the thing that moves.
  5. Keep the action simple. When the person runs, turns and waves while the camera orbits, the model has two moving things to keep straight. Let one lead: a still subject for a big camera move, a still camera for a big action.
  6. Describe the move, not the gear. "On a tripod" put a tripod in our pan. Say what the camera does ("stays in one spot", "does not move") and, if it matters, add that no camera equipment appears in the shot.

Watch for this:Read the clip before you use it. Compare the first and last frames: what got bigger, what turned, what slid past. A push in and a zoom end at the same size, but the background behaves differently; a pan and an orbit both turn, but only the orbit keeps the subject centered. Check the speed and the distance too: our crash zoom took about two seconds, and our orbit went less than the half circle we asked for.

Questions and answers

What is the difference between a dolly and a zoom?
A dolly moves the whole camera toward or away from the subject, so near things pass faster than far ones and the space feels deep. A zoom changes the lens while the camera stays put, so everything grows or shrinks evenly and the picture flattens. Write dolly or push in when the camera should travel, and zoom only when you mean the lens.
Can I use more than one camera move in a clip?
You can, but one move per shot comes back cleanest. If you need two, give them an order and a moment: a slow push in, then, after he sits down, a slow orbit to the right. Two short clips cut together usually look better.
Should I pick the move or write it?
Pick it from the Camera move row for one standard move: it adds a ready sentence at the strength you choose. Write it when the move needs a start, an end, a moment or a landing, or when the row does not have it, like a crash zoom out or bullet time. Do not do both.
Does a camera move cost extra?
No. Picking a move adds a sentence to your prompt, and the price of a clip comes from the model, the length and the quality. The Generate video button shows it before you start.
Why did the model do a different move?
A move is a request, and some read better than others. Our first bullet time came back as an ordinary throw until the prompt said time stops completely and nothing moves except the camera. Spell out what the camera does, what stays still and what the background does, then run it again: every clip is a new render.

Give your next clip a camera move

Open the video composer

Filmmaking

How to make realistic AI videos

The step-by-step workflow for realistic AI video: plan with the NOLGIA Agent, lock the characters and the place, write every shot beat by beat for Seedance 2.5, then score and cut.

· 20 min read

Video editing

Get every camera angle from one take

Turn one locked off take into a multicam edit: a real NOLGIA project from the take to the cut, with every prompt as it was sent, the angles that worked, the ones that missed and what it all took.

· 15 min read

Video editing

A clip came out wrong: fix it or make it again

Most wrong clips do not need a new render. Swap the one thing that is off, move the scene, clean up the picture or extend it, and make it again only when the shot itself is wrong.

· 5 min read