Seedance 2

Create cinematic text-to-video and reference-driven image-to-video clips in PixPark with Seedance 2 and Seedance 2 Fast.

Seedance 2.0 all-in-one references. Upload references, type text, or use @ subject for limitless creative possibilities.

Explore the Seedance video family

What Seedance Does Best

Prompt-only cinematic shots

Text only

Useful for storyboards, ad concepts, and mood pieces when you want clear motion and camera intent from text alone.

Blue-hour Tokyo overpass after rain, a courier on a silver motorbike cuts through shallow puddles, camera begins in a wide overhead shot and then drops into a low tracking angle beside the bike, taillight reflections streak across the wet asphalt, soft mist in the distance, subtle lens breathing, grounded cinematic realism, 6 seconds, 720p, 16:9.

Multi-reference fusion

Image + Video + Audio

Useful when one reference should define the look, another should define the motion, and audio should define timing or voice.

Keep the character face, silver jacket, and blue hair accents from Image 1. Borrow the shoulder-mounted handheld drift and slow push-in from Video 1. Use Audio 1 as the beat structure throughout. Night rooftop performance, the singer steps toward the camera, wind catches the jacket, city lights bloom in the background, and the final head turn lands exactly on the chorus hit. Clean continuity, high motion stability, 8 seconds, 720p, 16:9.

Edit existing footage

Video + Image

Useful for replacing subjects, changing backgrounds, or adding and removing objects while preserving the original shot.

Edit Video 1. Replace the ordinary black sedan with the retro-futuristic silver concept car from Image 1. Keep the road, camera path, vehicle speed, rain reflections, lighting direction, and all pedestrians unchanged. Preserve the original motion and timing, and make the new car feel naturally captured in the same shot. 6 seconds, 720p, 16:9.

Extend a scene naturally

Video extend

Useful when a strong shot ends too early and you want the action to continue without a visible reset.

Extend Video 1 forward by 4 seconds. Continue the rooftop dance after the final spin: the dancer lands, the coat hem settles in the wind, she takes one breath and looks over her shoulder, then the camera slowly cranes upward to reveal the city skyline behind her. Keep the same character identity, outfit, lighting, lens feel, and motion rhythm as Video 1, with no jump in continuity.

Native dialogue with synced audio

Audio-led

Useful for talking shots, music-led clips, and social scenes where voice timing matters as much as visuals.

Warm cafe interior at night, medium close-up. A young detective leans in and says, "I know you left before sunrise." Voice timbre references Audio 1. Let the line land naturally with believable lip sync, a soft pause before the final word, faint cup and spoon sounds in the background, and a subtle camera push-in during the sentence. 5 seconds, 720p, 16:9.

First-to-last frame control

2 frames

Useful when the opening composition and the ending payoff must be locked, not just suggested.

Start exactly from Image 1 as the first frame: a paper lantern floating in a dark river at dusk. End exactly on Image 2 as the last frame: the same lantern arriving near a bright festival gate in the distance. Between them, the camera glides low over the water, ripples widen, warm reflections multiply, and the lantern drifts steadily toward the gate. Keep the lantern shape, color, and continuity stable. 6 seconds, 720p, 16:9.

How to Prompt Seedance 2

Seedance works best when you clearly separate subject, motion, camera language, timing, and reference roles.

Prompt formula

[Scene] + [Main subject] + [Motion] + [Camera move] + [Lighting / mood] + [Reference role] + [Duration / resolution]

Start with the shot

Open with the environment, lens feeling, and composition so the model understands what kind of scene it is building.

Describe motion in verbs

Use verbs like drifts, tilts, lands, or pulls back instead of vague words like dynamic or cinematic.

Give references one job each

Use image references for look, video references for motion rhythm, and audio references for timing or beat alignment.

Lock continuity when needed

If identity, wardrobe, or frame layout must stay stable, say what must remain unchanged before adding new movement.

Keep timing explicit

Mention whether the motion should build slowly, hit on beat, or resolve in the last seconds of the clip.

Choose resolution and duration on purpose

Use 480p for faster ideation and 720p when motion detail, texture, and polish matter.

Quick example

Moonlit rooftop in Seoul, light rain, one dancer in a black coat stepping forward as the camera slowly pushes in. Keep the face and silhouette from the reference image, sync the final turn with the uploaded beat, and let the city bokeh stay soft in the background. 5 seconds, 720p, 16:9.

Don't

cool dance video cinematic city nice lighting dramatic motion

Seedance family comparison

Feature	Seedance 2	Seedance 2 Fast	Seedance 1 Pro
Best for	Highest-quality final shots	Faster iteration and prototyping	Cinematic text-to-video and image-to-video
Core inputs	Text, image, video, audio	Text, image, video, audio	Text, image
Native audio workflow	Yes	Yes	—
Reference depth	Image, video, audio, and mixed references	Same as Seedance 2	Prompt-led and image-led workflows
Video editing	Yes	Yes	—
Video extension	Yes	Yes	—
Visual control	Strongest overall control and fidelity	Same feature set with quicker turnaround	Strong prompt following and multi-shot storytelling
Resolution ceiling	Up to 720p	Up to 720p	Up to 1080p
Duration	4–15s	4–15s	Commonly shorter, provider-dependent
Aspect ratios	21:9, 16:9, 4:3, 1:1, 3:4, 9:16	21:9, 16:9, 4:3, 1:1, 3:4, 9:16	Multiple cinematic and social ratios
Workflow feel	Deliberate and polished	Quick and flexible	Cinematic and prompt-led
Choose it when	Motion detail, realism, and control matter most	You need more versions faster	You want text/image-driven cinematic clips without a multimodal audio/video workflow

Community feedback on Seedance

Seedance FAQ

What is the difference between Seedance 2 and Seedance 2 Fast?

Seedance 2 is the default higher-fidelity option, while Seedance 2 Fast is tuned for quicker iteration and faster turnaround.

Can I use image, video, and audio references together?

Yes. Seedance image-to-video supports mixed image, video, and audio references, as long as you provide at least one visual reference.

Does audio-only input work?

No. Audio references must be paired with an image or video reference so the model has a visual target.

How does pricing work for Seedance?

Video pricing depends on output duration, resolution, and whether you provide video references. Audio references do not change the billing rate.