Camera movement is the most underused element in Seedance 2.0 prompts. Most people describe subjects and scenes — what the shot contains — and skip how the camera sees it. The result is technically correct video that feels flat and static.
Seedance responds to camera instructions the same way it responds to scene instructions: with precision, when you write them precisely. Vague requests like "dramatic camera work" produce nothing useful. "Slow push-in from medium to extreme close-up, 15 seconds" produces exactly that.
This is the reference guide. Every major camera movement category, the exact wording that works, and a real prompt example for each.
Why Camera Movement Changes Everything
A scene described without camera instructions defaults to whatever Seedance calculates fits the content — usually a static medium shot. The same scene with a specific camera movement prompt becomes cinematic, active, and intentional.
Camera movement also solves the "dead frame" problem. In AI video, characters performing complex actions can freeze mid-motion if the prompt doesn't give the camera something to do. A tracking shot following a running character, a crane rising to reveal a landscape, a handheld shake during combat — these all keep the frame active throughout the full clip duration.
Category 1: Linear Camera Movements
These move the camera through space in a straight line. They're the most predictable and reliable in Seedance.
Dolly Push-In (Slow Push, Rack-In)
Moves the camera slowly toward the subject. Creates intimacy, tension, or revelation. Works well for emotional moments and product reveals.
Wording: slow push-in from medium to close-up / dolly in toward the subject / camera gradually moves closer, tightening on the face
See this technique in a Nutella product reveal →
"[0-5s] Smooth orbital shot circling a sealed jar of Nutella... [5-10s] Slow push-in to extreme close-up of the label..."
When to use it: Product reveals, character introductions, moments of decision or recognition. Combine with timeline splits — push-in works best as a second segment after an establishing wide shot.
Dolly Pull-Back (Reveal Shot)
Moves the camera backward from a subject to reveal context. The opposite of push-in — creates surprise, scale, or loneliness.
Wording: camera slowly pulls back to reveal the full environment / dolly out from close-up to wide, showing the scale of the surroundings / starts on face, pulls back to reveal the crowd
When to use it: Location reveals, scale contrast, isolation moments. If you start tight on a face and pull back to show a crowd or empty landscape, the contrast lands hard. Works best when the reveal is meaningful to the story.
Tracking Shot (Follow)
The camera moves alongside or behind a moving subject, keeping them in frame. Maintains energy and sense of motion.
Wording: tracking shot following the subject from behind / side tracking shot keeping pace with the running figure / camera tracks the character laterally as they move through the scene
When to use it: Action sequences, chase scenes, character walks. Combine with at shoulder height for an over-the-shoulder feel, or at knee height for a more aggressive perspective.
Category 2: Rotational and Arc Movements
These move the camera around a subject or point in space. They create dynamic, 3D-feeling shots that look expensive.
Orbital / Arc Shot
The camera circles around the subject while staying locked on them. One of the most cinematic moves available in Seedance.
Wording: smooth orbital shot circling the subject 180 degrees / camera arcs slowly around the figure, rotating 360 degrees / slow arc from left to right, keeping the subject centered
See a full 360-degree orbital shot in action →
"A dramatic intense scene, camera spin 360° around a male blond young adult, he is in pain as his bones dislocate and skin ripple, camera captures the intricate details as his body transforms gradually..."
When to use it: Transformations, product reveals (especially with a 180-degree arc), character reveals, and any moment that benefits from seeing a subject from multiple angles simultaneously. Orbital shots are self-contained — they work across the full clip length without needing timeline splits.
Crane Shot / Pedestal (Vertical Movement)
The camera moves vertically — rising (up) or descending (down) to change perspective. Creates a sense of scale and revelation.
Wording: crane up from ground level to aerial view / camera slowly rises, revealing the full cityscape below / pedestal down from above the scene to eye level
When to use it: Location reveals, moments of triumph (rising shot), descending into a scene (establishing shots for underground or interior locations). For aerial reveals, combine with a slight pull-back to maximize the scale effect.
Category 3: Handheld and Stabilized Movements
These describe how steady the camera is and imply a camera operator's physical presence in the scene.
Steadicam / Gimbal (Smooth Follow)
Floating, smooth movement that follows action without the mechanical feel of a dolly. Looks human-operated but without shakiness.
Wording: steadicam following the subject through the corridor / smooth gimbal shot as the camera weaves through the crowd / fluid handheld movement, no camera shake
See steadicam technique in a kitchen fight sequence →
"Camera Behavior: Continuous one-shot long take (steadicam), aggressive whip-pans, and rhythmic push-in zooms on impacts..."
When to use it: Action scenes where you want controlled motion rather than chaos. The "one-shot long take (steadicam)" instruction in the kitchen fight example makes Seedance commit to a single continuous movement — this prevents the default cut-heavy behavior.
Handheld / Documentary
The camera has intentional shake — like a human holding it without stabilization. Creates urgency, realism, and immediacy.
Wording: handheld camera, slight shakiness / documentary-style handheld, following the subject through the crowd / verité-style camera, unstabilized, ground-level
When to use it: Conflict scenes, intimate interviews, war or protest footage, any scene where "you are there" immediacy matters. Don't combine handheld with slow, deliberate subject movement — the contrast between a calm scene and shaky camera reads as unintentional rather than stylistic.
Category 4: Camera Angle and Position
These describe where the camera sits in space relative to the subject — not how it moves, but where it starts.
Low Angle
Camera positioned below eye level, looking up at the subject. Makes subjects appear powerful, threatening, or heroic.
Wording: low angle shot, camera at knee height looking up / worm's eye view, camera close to the ground pointing upward / Dutch low angle, camera tilted and low
When to use it: Heroes, villains, and any character whose authority you want to emphasize. Works especially well for the final moment of a revelation or arrival scene.
High Angle / Bird's Eye
Camera positioned above looking down. Makes subjects appear small, vulnerable, or surveilled.
Wording: overhead shot, camera pointing straight down / high angle looking down on the scene / bird's eye view from above, full environment visible
See overhead angle in an animation prompt →
When to use it: Crowds, maps, complex spatial relationships between characters, or any moment where the audience needs to understand the full geography of a scene.
Dutch Tilt
Camera rotated so the horizon is diagonal. Creates tension, instability, and psychological unease.
Wording: Dutch tilt, camera angled 20 degrees / tilted Dutch angle, slight rotation to the right / canted frame, horizon at 30-degree diagonal
When to use it: Villain introductions, psychological tension, scenes where the world feels "wrong." Don't overuse — one Dutch tilt per scene is effective; multiple in a row feels like a technical error.
Category 5: Optical Techniques
These control lens behavior rather than physical camera movement.
Zoom vs. Dolly
Zoom changes focal length; dolly physically moves the camera. They produce different effects — Seedance can do both.
Wording for zoom: slow zoom in on the subject's face / rapid crash zoom at the moment of impact / optical zoom from wide to telephoto
Wording for dolly: dolly in as opposed to a zoom, camera physically moves forward
When to use each: Crash zoom is a classic stylistic move for comedic or dramatic emphasis. A slow zoom feels slightly "zoomed in" — the background compression increases as you zoom. Dolly maintains natural perspective distortion.
Rack Focus
The focus plane shifts from one subject to another. Draws attention to different parts of the scene.
Wording: rack focus from foreground character to background subject / shallow depth of field, focus shifts from near to far / soft foreground, sharp background — then focus reverses
When to use it: Two-character scenes, moments of revelation where the "important" element isn't the subject the audience expects. Works best with a shallow depth of field (shallow DoF or wide aperture, f/1.4).
Combining Multiple Camera Instructions
Single camera instructions work. Combining two or three produces the most cinematic results.
Formula: [movement] + [position/angle] + [lens/optical]
Examples:
low angle tracking shot following from behind, shallow depth of fieldsmooth orbital arc, camera rising slowly while circling, wide angle lenssteadicam follow through the corridor, push-in as it approaches the door, 24mmhandheld documentary-style, low angle, with rack focus shift mid-shot
Place camera instructions in their own line or block within the prompt — mixing them into subject descriptions creates ambiguity.
The Continuous Move Instruction
One of the most powerful single camera instructions for Seedance:
CONTINUOUS MOVE UNTIL [event]
This prevents Seedance from cutting to a new angle mid-clip. The model defaults to editing behavior — it will cut to closer shots, wider shots, or different angles if not instructed otherwise. "CONTINUOUS MOVE UNTIL MATCH CUT" in the VR survival prompt below forces the model to maintain a single unbroken camera movement until the specified transition point.
See the continuous move instruction in a VR action sequence →
"CONTINUOUS MOVE UNTIL MATCH CUT + IMMEDIATE ACTION FROM FIRST FRAME"
Use CONTINUOUS MOVE when you want a long take, a one-shot sequence, or any scene where cutting would break the effect.
Frequently Asked Questions
What camera movement instructions does Seedance 2.0 actually respond to?
Seedance responds well to film-industry terminology: push-in, pull-back, tracking shot, orbital shot, crane up, steadicam, handheld, Dutch tilt, rack focus. Terms like "dramatic camera" or "movie-style camera work" are too vague — Seedance defaults to static medium shots when instructions aren't specific.
How do I stop Seedance from cutting to multiple angles when I want a single shot?
Add CONTINUOUS MOVE or continuous one-shot take to your prompt. Without it, Seedance treats long clips as multi-shot sequences and edits them accordingly.
Where in the prompt should I put camera instructions?
Two approaches work: (1) in a dedicated FORMAT or [CINEMATIC SETUP] block at the top of the prompt, or (2) in each timeline segment individually. The top-of-prompt placement sets the global camera behavior; per-segment placement overrides it for that segment.
Can I specify both camera movement and camera angle in the same prompt?
Yes — and you should. Combining position (low angle), movement (tracking), and lens (shallow depth of field) produces results that match how real cinematographers describe shots. Keep camera instructions grouped together rather than scattered through the prompt.
Does camera movement affect audio output in Seedance?
Not directly, but the overall cinematographic register you establish often carries into audio. A handheld action sequence with a crash zoom tends to produce more urgent, percussive audio than a slow crane shot — the visual language bleeds into the audio simulation.
What's the difference between a zoom and a dolly in AI video?
In real cinematography: a zoom changes focal length (background compression increases), a dolly physically moves the camera (perspective stays natural). Seedance simulates both but responds better to physical movement terms. If your result looks "zoomed" when you want a natural push-in, switch from zoom in to dolly in / slow push-in.
Start Here
If you're adding camera movement to your prompts for the first time:
- Start with orbital shots — they're the most reliable and cinematic in Seedance
- Add the
CONTINUOUS MOVEinstruction to any clip over 8 seconds - Combine a position (low angle) with a movement (tracking) for richer results
- Use timeline splits to change camera movement between segments
Browse prompts on scenic.sh to see how camera instructions affect real output videos — each prompt card shows the full text alongside the result.
