Seedance 2.0 Prompts: A Practical Prompt Guide

Good Seedance 2.0 prompts read like a short shot list, not a poem. Seedance 2.0 is ByteDance Seed's video model: it takes text, images, video and audio as input, generates the picture and the sound together, and can cut between several shots inside one clip of up to 15 seconds (ByteDance Seed). This Seedance 2.0 prompt guide covers the structure we use, how to time several shots, how to point at your uploads, and 12 prompts you can paste into the generator as they are.

Everything here also works with Seedance 2.5, which takes the same kind of prompt but allows up to 30 seconds per clip.

The Seedance 2.0 prompt structure

Write the prompt in this order. Not every prompt needs every part.

Subject. Who or what, with two or three visual details. "A woman in a yellow raincoat" beats "a woman".
Action. What happens, with a verb you could film. One main action per shot.
Camera. Shot size and movement: close-up, wide shot, slow push-in, handheld follow, orbit, locked-off tripod.
Light and style. Time of day, light source and look: "golden-hour backlight", "neon night", "35mm film grain", "clean studio product shot".
Sound. Seedance makes audio with the picture, so say what should be heard: ambience, effects, a music genre, or dialogue.

Put together, a short prompt looks like this:

A woman in a yellow raincoat runs across a wet city street at night, handheld camera following at her shoulder, neon reflections on the asphalt. Sound: rain and passing traffic, no music.

Put dialogue in quotes

Write spoken lines in quotation marks and say who speaks and how: The barista smiles and says, "Your usual? It's on the house today." A 5-second shot holds about one sentence.

Say what you don't want, briefly

Short, plain instructions such as "no music", "no subtitles" or "one continuous shot, no cuts" work better than a long list of negatives.

Timed beats for multi-shot Seedance prompts

Seedance 2.0 can put several shots in one clip. When the order matters, write timed beats:

0–4 s: Wide shot of a lighthouse on a cliff at dawn, waves crashing below.
4–9 s: Medium shot, the keeper in a grey coat climbs the spiral stairs with a lantern, footsteps echoing.
9–15 s: Close-up of the lamp lighting up, then the camera pulls back through the window to the open sea. Low strings swell.

A few rules keep beats working:

Make the beats add up to the length you choose in the form (15 s here).
Give each beat one shot and one action. Three or four beats are plenty for 15 seconds.
Repeat the key detail ("the keeper in a grey coat") when a subject comes back, so it stays the same person.
Want one unbroken take instead? Write "one continuous shot, no cuts" and describe the camera path from start to end.

How to reference Image 1, Video 1 and Audio 1

In Reference mode on this site you can add up to 9 images, 3 video clips (2–15 s each, 15 s in total) and 3 audio files (15 s in total). Each upload gets a name, such as Image 1, Image 2, Video 1 or Audio 1, and typing @ in the prompt inserts it. References guide the video; they are not used as its frames.

The trick is to tell the model what job each reference does:

Images set identity, outfit, product or style: "the character from Image 1", "the sneaker in Image 2", "the color palette of Image 3".
Videos set motion and camera: "the dance moves from Video 1", "follow the camera move in Video 1".
Audio sets music or voice: "Audio 1 as the soundtrack", "she speaks in the voice from Audio 1".

Audio needs at least one image or video next to it. Reference video seconds are billed like output seconds, so trim clips to the part you need. The reference-to-video page has more on this mode.

Image-to-video prompts: describe the motion, not the picture

In Image mode your picture becomes the first frame, and you can add a second picture as the last frame. The model already sees the image, so describe what changes:

Weak: "A girl with red hair in a green dress standing in a field of flowers."
Better: "She turns toward the camera and laughs as the wind lifts her hair; the camera slowly pushes in. Birdsong and a light breeze."

With a last frame, describe the path between the two. The video takes the shape of your first image, so crop it first. See image-to-video for examples.

12 copy-ready Seedance 2.0 prompts

Paste these into /create and swap in your own subjects. Mode, aspect ratio and length are in parentheses.

Cinematic

1. Harbor at dawn (Text, 16:9, 10 s)

0–5 s: Wide shot of an old fishing boat leaving a misty harbor at dawn, gulls circling overhead.
5–10 s: Close-up of the weathered captain at the wheel, squinting into the fog.
Muted blue-grey palette, 35mm film grain. Sound: creaking wood, gulls, a low foghorn, no music.

2. Desert rider (Text, 21:9, 8 s)

One continuous shot. The camera glides low over a desert road at sunset toward a lone motorcyclist, then rises and circles as she stops and lifts her visor. Warm orange backlight, long shadows, anamorphic lens flare. Sound: engine idling, wind, a slow synth pad.

3. Rain noir (Text, 16:9, 15 s)

0–5 s: Night, heavy rain, wide shot of a detective in a long coat walking under a flickering streetlight.
5–10 s: Medium shot, he stops at a phone booth as the phone starts ringing.
10–15 s: Close-up, he picks up and says quietly, "I know it was you."
Black-and-white film noir, high contrast. Sound: rain, distant sirens, the phone ringing.

Product

4. Earbuds reveal (Text, 1:1, 6 s)

A matte black earbud case slowly rotates on a white pedestal; the lid opens and the earbuds lift slightly with a soft glow. Clean studio lighting, soft shadows, macro close-up. Sound: a subtle click and a soft electronic chime. No text on screen.

5. Perfume on stone (Reference with one product image, 16:9, 8 s)

The perfume bottle from Image 1 stands on wet black stone; water drops fall in slow motion and splash around it while the camera slowly circles. Dark, moody light with a single warm rim light. Sound: water drops, one low cello note.

Dance and motion

6. Rooftop routine (Reference: Image 1 dancer, Video 1 moves, Audio 1 music, 9:16, 10 s)

The dancer from Image 1 performs the moves from Video 1 on a rooftop at sunset, full body in frame, static camera at chest height. Audio 1 as the music.

7. Under the bridge (Text, 9:16, 10 s)

0–5 s: Wide shot of a street dancer in a red hoodie popping and locking under a bridge.
5–10 s: Close-up of his sneakers sliding into a moonwalk, then the camera tilts up to his grin.
Sound: a punchy hip-hop beat, a small crowd cheering.

Dialogue

8. Park bench (Text, 16:9, 8 s)

Two friends on a park bench in autumn, medium two-shot. The woman says, "You actually came back." The man smiles and answers, "I said I would." Soft afternoon light, falling leaves. Sound: rustling leaves, distant children playing, no music.

9. Travel vlog intro (Image, your selfie, 9:16, 6 s)

She looks into the camera and says cheerfully, "Three things I pack for every trip. Let's go." Natural vlog look, handheld with slight shake. Sound: her voice and light room ambience.

Ads in 9:16

10. Energy drink turn (Text, 9:16, 10 s)

0–3 s: Close-up of a tired office worker yawning at her desk under flat fluorescent light.
3–7 s: She opens a can of sparkling lemon drink; bubbles rush up and the room turns bright and warm.
7–10 s: She grins at the camera and says, "Better."
Sound: can opening, fizz, upbeat pop music.

11. Sneaker drop (Reference with one product image, 9:16, 8 s)

The sneaker from Image 1 drops onto a wet basketball court in slow motion and splashes water; the camera circles low around it, then a player's foot slides in and the laces pull tight. Night, stadium lights. Sound: a heavy thud, a splash, a bass hit.

12. Poke bowl build (Text, 9:16, 6 s)

Top-down shot of hands building a poke bowl: rice, salmon, avocado and sesame land one after another in quick rhythmic cuts. Bright natural light, vivid colors. Sound: crisp chopping and plating, a light upbeat beat. No text on screen.

Common mistakes with Seedance prompts

Too much in one shot. Five actions in 5 seconds give you a blur. Split them into beats or pick one.
Describing the picture in Image mode. The model sees it. Spend the words on motion, camera and sound.
Beats that don't match the length. Beats written for 15 s on a 5 s render get squeezed or dropped.
"Cinematic" as the only camera note. Name the shot and the move: "slow dolly-in, close-up".
Forgetting sound. If you don't say, the model picks for you.
Unnamed references. "Use the photo" is ambiguous with three uploads; "the jacket from Image 2" is not.
People who didn't agree. ByteDance Seed requires identity verification or prior authorization to use real people's portraits as subject references. On this site, only upload people who agreed to appear, and no one under 16.

FAQ

How long should a Seedance 2.0 prompt be?

Usually two to five sentences, or one line per beat for multi-shot clips. Repeating the same idea in more words doesn't help.

Do Seedance 2.0 prompts work in Seedance 2.5?

Yes. The structure is the same; Seedance 2.5 just gives you up to 30 seconds, so you can write more beats. Seedance 2.0 and 2.0 Fast stop at 15 seconds.

Why did my clip get music I didn't ask for?

Sound is generated with the picture, and without directions the model may add a score. Write "Sound: … no music" with the effects or ambience you do want.

Can I test these prompts for free?

New accounts get free credits for one 5-second video, and the daily check-in gives one free Seedance 2.0 Fast 480P clip of up to 5 seconds in Text or Image mode. See Seedance 2.0 free and pricing. For a walkthrough of the form, read how to use Seedance 2.0.

Seedance is a trademark of ByteDance. This site is an independent service and is not affiliated with or endorsed by ByteDance.