Seedance 2.0 Prompting Guide (2026)

This single prompt gives you a werewolf transformation. With a 360-degree rotating camera:

"A dynamic 360-degree rotating camera captures the transformation of a man into a werewolf under the full moon..."

This prompt makes a 3-segment Nutella commercial:

"[0-5s] Start with a smooth orbital shot circling a sealed glass jar of Nutella... [5-10s] Cut to extreme close-up of a silver knife slowly dragging a thick layer of Nutella across a perfectly toasted slice of bread..."

See the result video →

What makes these different from a prompt that just says "cinematic werewolf transformation"? Structure. This guide breaks down that structure end to end.

The 4-Layer Formula

Working Seedance prompts have 4 layers. Drop one and the model fills in the blank — almost always in the wrong direction.

1. Subject — Who or what is in the frame

Be specific. "A woman" gives the model too much freedom. "A tired suburban mom in yoga pants pushing an oversized flatbed cart" is a clear instruction.

Bad: A person walking Good: A young woman in a flowing white dress walking barefoot Better: A weathered fisherman in a salt-stained yellow raincoat hauling a net over his shoulder

2. Action — What they're doing

One action per segment. Write "She picks up a cup, walks to the window, and looks outside" and the model will skip one.

Bad: She dances and then sits down and drinks tea Good: She reaches for a porcelain teacup with both hands, steam curling upward

3. Environment — Where, when, what it feels like

Lighting, weather, time of day, surrounding detail. Most prompts skip this — they describe subject and action and forget the world.

"warm morning sunlight streaming through a window creating golden highlights and soft shadows on the glossy jar label"

That single line in the Nutella prompt does more for output quality than three lines describing the subject.

4. Camera & Style — How the audience sees it

Camera angle, movement, film style, mood. This is where the multiplier lives.

tracking shot following from behind — action sequences
slow push-in from medium to close-up — emotional moments
smooth orbital shot circling — product reveals
handheld, slight shake — documentary feel
static wide shot — scene-setting

More camera-movement examples →

Timeline Splitting — The #1 Technique

If your video is over 5 seconds, split the timeline. This single technique separates amateur prompts from pro prompts.

How

[0-5s] First action — opening shot, reveal, or hook
[5-10s] Second action — development, close-up, or transition
[10-15s] Third action — climax or resolution

Real example

See this prompt on scenic.sh →

"[0-5s] Start with a smooth orbital shot circling a sealed glass jar of Nutella centered on a rustic wooden table, warm morning sunlight streaming through a window..."

"[5-10s] Cut to extreme close-up of a silver knife slowly dragging a thick layer of Nutella across a perfectly toasted slice of bread..."

Rules:

3-second intervals as default unit
One core action per segment — never two
The end state of one segment should flow naturally into the start of the next
10s video: 2-3 segments. 15s video: 4-5 segments.

Without timeline splits, Seedance picks the pacing on its own. Almost always rushes through the good parts.

Deep dive on this technique →

Camera Moves That Actually Work

Camera direction is half the prompt. Here are the moves Seedance 2.0 handles best, sorted by reliability:

Tier 1 — Almost always works

Move	Use for	Phrasing
Slow push-in	Emotional reveal	`slow push-in from medium shot to extreme close-up`
Orbital / circling	Product shots, character intros	`smooth 360-degree orbital shot circling the subject`
Tracking (behind)	Chase scenes, follow-the-action	`tracking shot following from behind at shoulder height`

Tier 2 — Works when context is clear

Move	Use for	Phrasing
Crane up/down	Scale reveal	`crane shot rising from ground level to reveal the city below`
Dolly zoom	Tension, vertigo	`dolly zoom creating a vertigo effect`
Whip pan	Scene transition	`whip pan to the right transitioning to the next scene`

Tier 3 — Use carefully

Move	Use for	Phrasing
First-person POV	VR, horror, exploration	`first-person POV camera pushing through the door`
Match cut	Advanced transition	`match cut from the spinning wheel to the spinning planet`

More camera prompts →

Style Modifiers — Setting Visual Tone

Adding style modifiers at the start or end of a prompt completely changes the result. The ones that work consistently:

Film styles

Shot on 35mm film, warm color grading — vintage cinematic
IMAX footage, ultra-wide aspect ratio — epic scale
Shot on iPhone, vertical format — social-media realism
8mm home video, light leaks and grain — nostalgia

Lighting keywords

golden hour lighting — warm, soft
neon-lit, cyberpunk lighting — sci-fi
harsh overhead fluorescent — tension, institutional
volumetric fog with backlight — atmospheric, dramatic

Example

See this prompt on scenic.sh →

This VR action prompt stacks several style modifiers: first-person POV + arctic lighting + survival horror tension. Style words don't describe — they tell the model what kind of film this is.

5 cinematic prompts that look like real movies →

Reference Materials — The @ Syntax

Seedance 2.0 supports quad-modal input — text, image, video, audio. When uploading reference materials, use @ syntax to declare what they're for.

Image references

Upload a reference image and tag it in the prompt:

@image1: character appearance and clothing
A young woman wearing the outfit from @image1 walks through a neon-lit Tokyo alley at night.
Tracking shot from behind, rain-slicked pavement reflecting pink and blue signs.

To keep a character consistent across multiple shots, add this phrase: maintain exact appearance from reference image, consistent character throughout, no deformation or drift

Audio-synced video

Upload an audio track and the model edits to the beat — beat-matched cuts, lip-sync, emotion-matched pacing. Especially powerful for music videos and montages.

@audio1: background music track
A dancer in a red dress performs contemporary dance in an empty warehouse.
Movements synchronized to the beat of @audio1, dramatic lighting shifts on each drop.

Prompt Length: How Far Is Too Far?

After testing hundreds of prompts on scenic.sh:

Length	Best for	Risk
1-2 sentences	Simple scenes, static shots	Model fills in too many blanks
3-5 sentences	Most cases	Sweet spot
6-10 sentences	Timeline splits, complex scenes	Works if structured well
10+ sentences	Multi-segment ads	Tail end may be ignored

The key isn't length — it's information density. Every sentence should add something the model can act on. Drop adjectives that don't change the visual outcome.

"A beautiful, stunning, gorgeous sunset over the ocean" → same result as "sunset over the ocean."

"A sunset over the ocean, the water reflecting burnt orange, a single fishing boat silhouetted against the sky" → completely different result.

Common Mistakes

1. Cramming too many actions into one segment

The model can handle one smooth action per 3-5 seconds. Ask for more and something gets dropped.

2. Vague camera direction

"Cinematic camera movement" means nothing. "Slow tracking shot moving left to right at eye level" gives the model a clear job.

3. Physically impossible directions

"A wide shot of a person's face in extreme close-up" — the model can't do both. Pick one framing per segment.

4. Missing environment

Prompts where the subject just performs an action in empty space. Always anchor the scene to a specific location with specific lighting.

5. "Cinematic" overuse

The word "cinematic" alone won't make your video cinematic. What makes it cinematic is specific camera moves, intentional lighting, meaningful framing. Write those instead.

Quick Start Templates

Person / character

A [specific character description] in a [specific location].
[Lighting description]. [Camera move] capturing [what's visible].
[Mood/style modifier].

Example: A weathered fisherman in a salt-stained yellow raincoat standing at the bow of a wooden boat. Dawn light breaking through storm clouds, catching the spray. Slow push-in from medium shot, 35mm film grain.

Product / commercial

[0-3s] [Product reveal shot — orbital or push-in, dramatic lighting]
[3-7s] [Product in use — close-up, human interaction]
[7-10s] [Lifestyle / emotional payoff — pull back to show context]

Action / dynamic

[Specific choreographed action description].
[Camera: tracking/handheld/specific move].
[Speed: slow-motion, match cut, or real-time].
[Environment anchor — where it happens].

Browse 1000+ working prompt examples →

By Category

If you're looking for a specific style:

Action prompts — chases, combat, explosions
Cinematic prompts — film-grade shots and lighting
Animation prompts — cartoon and animation styles
Anime prompts — Japanese animation aesthetics
Slow-motion prompts — dramatic time manipulation
Aerial prompts — drone and bird's-eye views
Nature prompts — landscapes, weather, wildlife
Portrait prompts — character-centric shots
Food prompts — culinary cinematography