Seedance prompts for ads

How do you write a Seedance prompt?

Write a Seedance prompt as an instruction, not as ad copy: who, where, doing what, how the camera moves, and in what order. BytePlus's Seedance 2.0 prompt guide gives the order as precise subject, action details, scene, lighting and color tone, camera movement, visual style, image quality, then constraints.

The guide adds three rules that shape every template on this page:

Inside each shot, the guide's order is camera move or transition, subject action and expression, position in the scene, then audio.

What syntax does Seedance use for dialogue and sound?

Seedance uses symbols to separate information types. The guide lists curly braces for dialogue, angle brackets for sound effects, parentheses for music, and a separate bracket style for subtitles.

• Dialogue: {This is the line}, after the speaker.
• Sound effect: <a kettle clicks off>.
• Music: (soft acoustic guitar in the background).

Keep each line in one language. Avoid per-shot timings such as 0 to 3 seconds: BytePlus says precise timing support is unstable. Duration (4 to 15 seconds on the 2.0 series) and aspect ratio are set as parameters in the video generation API, not in the prompt.

The templates below are written for this page and have not been run through Seedance; no outputs are shown. Square-bracket items are placeholders. For the specs and the full ad workflow, see Seedance for ads. To fill a template from a form, use the AI ad prompt builder.

Seedance prompt for a UGC selfie ad

A UGC selfie prompt needs one presenter label, one product label and three short shots. Compare the finished look with the get-ready UGC examples.

Define the [age range] woman with [hair] and [top] in Image 1 as Creator.
Define the [product colour] [product type] in Image 2 as Product.

Shot 1: Fixed front-camera selfie framing at arm's length. Creator sits on
[bed edge / car seat], leans slightly toward the lens, eyebrows raised.
She says {[hook line, under 12 words]}. <faint traffic outside>

Shot 2: Cut to a close-up of Creator's hand lifting Product into frame,
turning it slowly so the label faces the lens.
She says {[one benefit, one sentence]}.

Shot 3: Back to selfie framing. Creator smiles, nods once and says
{[call to action]}.

Natural phone-camera look, soft window light, realistic skin texture.
Keep it subtitle-free. Do not generate a logo or watermark.

Seedance prompt for a podcast clip ad

For a podcast clip, label both hosts from separate reference images and alternate shots between them. See the AI podcast ad examples for framing.

Define the man with [beard / glasses] in Image 1 as Host.
Define the woman with [hair] in Image 2 as Guest.
Use Image 3 as the studio scene reference.

Shot 1: Fixed medium shot of Host at a microphone, headphones on,
looking toward Guest off-camera. Host asks {[question that sets up the problem]}.

Shot 2: Cut to a fixed medium shot of Guest. She pauses, tilts her head,
then answers {[answer that names the product]}.

Shot 3: Slow push-in on Guest as she holds up [product] beside the mic.
Guest says {[one-line proof or detail]}.

Warm practical lamps, shallow depth of field, room tone only.
Keep it subtitle-free. Do not generate a logo.

Seedance prompt for a street interview ad

A street interview needs an interviewer label even if they stay off screen, plus a passer-by. The AI street interview examples show the handheld look.

Define the [age range] man in [jacket] in Image 1 as Passer-by.
The interviewer stays off camera; only a handheld microphone is visible.

Shot 1: Handheld medium shot on a busy [city] pavement in daylight.
The microphone enters frame. Interviewer asks {[quick question]}.
<street noise, distant bus>

Shot 2: Same framing. Passer-by laughs, glances at [product] in his hand
and answers {[short answer with the product name]}.

Shot 3: Slow push-in to a close-up as Passer-by holds [product] toward the lens
and says {[final line]}.

Documentary style, natural light, slight realistic camera shake.
Keep it subtitle-free. Do not generate a logo or watermark.

Seedance prompt for a short drama ad

Short drama is where the guide's emotion rule matters most: show the feeling through hands, shoulders and breath. See the AI drama ad examples.

Define the woman in [outfit] in Image 1 as Lead.
Define the man in [outfit] in Image 2 as Partner.
Use Image 3 as the [kitchen / office] scene reference.

Shot 1: Fixed wide shot. Lead stands at the counter, fingers tapping the
surface, eyes on [problem object]. Partner walks in slowly.
Partner asks {[question that names the problem]}.

Shot 2: Cut to a close-up of Lead. Her shoulders drop, she lets out a long
breath and says {[line admitting the problem]}.

Shot 3: Slow pull-back to a two-shot. Partner places [product] on the counter.
Lead picks it up, the corners of her mouth rise, and she says {[payoff line]}.
(soft piano, rising)

Cinematic look, warm tones, soft light. Faces stay stable, movements natural.
Keep it subtitle-free.

Seedance prompt for a claymation or animation ad

Name the style in the prompt. BytePlus warns that a realistic reference image can pull the output into live action unless you add explicit style words. Compare with the claymation and craft examples or the Pixar-style examples.

Stop-motion claymation style, handmade clay textures, visible fingerprints.
Define the clay figure of [product] with eyes and small arms in Image 1 as Mascot.

Shot 1: Fixed wide shot of a clay [bathroom shelf / kitchen counter].
Mascot slowly slides out from behind [problem object] and looks around.
<soft squelch of clay>

Shot 2: Cut to a medium shot. Mascot turns to the camera, waves one arm
and says {[one-line problem]}.

Shot 3: Slow push-in as Mascot jumps gently onto [surface] and points at
[benefit visual]. Mascot says {[one-line benefit]}.

Bright, soft studio light, [colour palette]. Keep the clay style throughout.
Keep it subtitle-free.

Seedance prompt for a product shot ad

Product shots use one subject and one camera move per shot, with the brand line as text. Seedance can render slogans; the guide's text template is text content, timing, position, entrance style and visual attributes. See the AI product demo examples.

Define the [product] in Image 1 as Product.

Shot 1: Macro close-up, fixed camera. A drop of [liquid / water] runs down
Product's surface under a hard side light. <single drip>

Shot 2: Smooth lateral tracking shot along Product on a [stone / wood] surface,
label facing the lens.

Shot 3: Slow push-in to a centred front view of Product. At the end, the text
"[brand line]" fades in at the top centre in [colour] [sans-serif] letters.

Commercial look, rich detail, clean background in [colour].
Do not generate any other text, logo or watermark.

Seedance prompt for a brand film ad

Brand films need the most atmosphere and the fewest words. Keep each shot to one move, and let sound carry the mood. See the AI brand film examples.

Define the woman with [hair] in [coat] in Image 1 as Runner.
Use Image 2 as the [coastline / city] scene reference.

Shot 1: Aerial wide shot drifting forward over [location] at dawn,
low mist, Runner small in the frame. <wind>

Shot 2: Cut to a side tracking shot at ground level. Runner jogs at an
even pace, breath visible, [product] on her wrist.

Shot 3: Fixed close-up. Runner stops, rests her hands on her knees, looks up
toward the horizon and smiles faintly.
(slow strings build, then stop)

Film grain, cool tones, low saturation. No dialogue.
Keep it subtitle-free. Do not generate a logo or watermark.

How do you join Seedance clips into a full ad?

Generate each script beat as its own clip and cut them together in an editor. BytePlus recommends video extension for one continuous dialogue scene, and separate clips stitched together for plot turns and fast action.

When you extend a clip, the guide suggests trimming 6 frames from the end of the earlier clip and 1 frame from the start of the next to hide the join. For more on running one script in every format, read one script, every format. If your Seedance ads already perform on paid traffic, see how Anglr works for creators.