All posts
Prompt guidesVideo to Prompt Team

Seedance 2.0 Prompt Guide (with Copy-Paste Prompts)

How to write Seedance 2.0 prompts — the prompt formula, a camera-language glossary, 22 copy-paste prompts by scenario and fixes for common mistakes.

Seedance 2.0 Prompt Guide (with Copy-Paste Prompts)

Seedance 2.0 is ByteDance Seed's video model, launched on February 12, 2026. It takes text, images, video clips and audio as input and returns a short clip with sound. That flexibility is also why prompts go wrong: you can describe almost anything, so it is easy to describe too much. This guide gives you a formula, a camera vocabulary, 22 prompts you can paste straight in, and the mistakes that cost the most retries.

If you would rather start from a rough idea and get a structured draft back, use our free seedance prompt generator, then come back here to polish it.

What Seedance 2.0 can do (the short version)

Everything in this section comes from ByteDance, BytePlus, Dreamina or CapCut; links are in the sources list at the end.

  • Inputs: text, images, video and audio. ByteDance Seed says you can combine up to 9 images, 3 video clips and 3 audio clips with a text instruction.
  • Modes: text-to-video, image-to-video from a first frame or a first and last frame, reference-based generation, video editing and video extension.
  • Sound: audio is generated with the picture, including voices, sound effects and background music.
  • Length and shape: 4–15 seconds per generation, at 24 fps, in 21:9, 16:9, 4:3, 1:1, 3:4 or 9:16.
  • Variants: BytePlus lists Seedance 2.0, Seedance 2.0 Fast and Seedance 2.0 Mini, which trade quality against speed and cost. Resolution options depend on the variant and platform.
  • Faces: reference images or videos of real human faces cannot be uploaded directly.

The Seedance 2.0 prompt formula

BytePlus's official prompt guide describes a good Seedance 2.0 prompt as an "engineering-style instruction", not ad copy. Its advanced formula is:

precise subject + action details + scene/environment + lighting & color tone + camera movement + visual style + image quality + constraints

For everyday use, you can compress that into four core parts plus three optional ones:

  1. Subject — who or what, with two or three stable details (clothing, color, material).
  2. Action — what the subject does, broken into small, specific movements.
  3. Camera — shot size, angle and one camera movement.
  4. Style — the look: realism, film stock, animation style, color tone.
  5. Optional lighting — time of day, light source, contrast.
  6. Optional audio — dialogue, sound effects, music. Seedance 2.0 generates sound by default, so leaving it out means the model decides.
  7. Optional duration note — pick the length in the duration setting (4–15 seconds). The official guide warns that exact timings inside a prompt, such as "0–3 seconds", are unreliable, so describe the order of events instead.

A worked example, line by line

Start with the subject:

A ceramic coffee mug with a chipped blue glaze sits on a wooden windowsill.

Add the action. Keep it slow and physical:

A ceramic coffee mug with a chipped blue glaze sits on a wooden windowsill. Steam rises slowly from the coffee, curling in the light breeze from the open window.

Add the camera — one shot size, one move:

A ceramic coffee mug with a chipped blue glaze sits on a wooden windowsill. Steam rises slowly from the coffee, curling in the light breeze from the open window. Close-up at eye level, slow push-in toward the rim of the mug.

Add style and lighting:

A ceramic coffee mug with a chipped blue glaze sits on a wooden windowsill. Steam rises slowly from the coffee, curling in the light breeze from the open window. Close-up at eye level, slow push-in toward the rim of the mug. Early morning sunlight from the left, soft shadows, warm tones, shallow depth of field, cinematic texture.

Finish with audio and constraints:

A ceramic coffee mug with a chipped blue glaze sits on a wooden windowsill. Steam rises slowly from the coffee, curling in the light breeze from the open window. Close-up at eye level, slow push-in toward the rim of the mug. Early morning sunlight from the left, soft shadows, warm tones, shallow depth of field, cinematic texture. Sound: distant birdsong and a faint kettle click. Keep it subtitle-free, no logo, no watermark.

That final version is about 70 words. Every sentence tells the model something it cannot guess, and nothing contradicts anything else.

Camera language glossary for Seedance

The official guide says Seedance 2.0 understands standard camera terminology, and recommends one camera movement per shot. Use the phrases below as they are, or swap in your own subject.

Shot sizes

TermWhat it meansSeedance-ready phrase
Extreme wide shotThe setting dominates; people are tinyextreme wide shot of a lone hiker on a salt flat
Wide shotWhole subject plus plenty of surroundingswide shot of the market street at dusk
Full shotHead to toe, little extra spacefull shot of the dancer in the empty studio
Medium shotWaist up; good for gesturesmedium shot of the barista behind the counter
Medium close-upChest up; good for dialoguemedium close-up of the detective as she speaks
Close-upFace or one object fills the frameclose-up of his hands tying the knot
Extreme close-up / macroA tiny detail, magnifiedmacro shot of a dew drop on a leaf

Camera angles

TermWhat it meansSeedance-ready phrase
Eye levelNeutral, at the subject's eye heighteye-level shot, steady framing
Low angleCamera looks up; the subject feels powerfullow-angle shot looking up at the statue
High angleCamera looks down; the subject feels smallhigh-angle shot of the child in the hallway
Overhead / top-downStraight down from abovetop-down shot of the chef plating the dish
Aerial / droneHigh above, often movingaerial drone shot over the coastline
Over-the-shoulderBehind one person, facing anotherover-the-shoulder shot from behind the father
POVWhat the character seesfirst-person POV walking through the forest

Camera movements

TermWhat it meansSeedance-ready phrase
Fixed / staticCamera does not movefixed camera position, locked-off shot
Push-in (dolly in)Camera moves toward the subjectslow push-in toward her face
Pull-out (dolly out)Camera moves away, revealing morecamera slowly pulls back to reveal the room
PanCamera turns left or right on the spotcamera pans right across the skyline
TiltCamera turns up or down on the spotcamera tilts up from his boots to his face
Truck / lateral trackingCamera slides sideways alongside the actionsmooth lateral tracking shot alongside the runner
Tracking / followCamera follows a moving subjectcamera follows the cyclist from behind
Orbit (circle around)Camera circles the subjectcamera slowly circles around the sculpture
Crane / riseCamera rises or descends verticallycamera rises above the crowd
HandheldSlight natural shake, documentary feelhandheld camera with slight realistic shake
ZoomLens magnification changes; camera stays putslow zoom in on the clock face

Lens and focus

TermWhat it meansSeedance-ready phrase
Shallow depth of fieldSubject sharp, background blurredshallow depth of field, soft background bokeh
Deep focusEverything sharp, front to backdeep focus, foreground and background both sharp
Rack focusFocus shifts from one subject to anotherfocus shifts from the glass to the woman behind it
Wide-angle lensExaggerated space, close to the actionwide-angle lens, close to the skateboarder
TelephotoCompressed background, distant feeltelephoto shot compressing the city behind him
Foreground framingSomething in front frames the subjectshot through the leaves in the foreground

22 copy-paste Seedance 2.0 prompts

Every prompt below is original. Paste it in, then change the subject, product or place. Dialogue sits in quotation marks, which BytePlus recommends for spoken lines.

Product ads

1. Skincare bottle reveal

A frosted glass serum bottle with a gold dropper stands on a white marble ledge. A single drop slides down the side of the bottle and catches the light. Slow orbit around the bottle at eye level. Soft studio lighting, pale peach background, clean minimal commercial style, rich detail. Sound: a gentle water droplet and soft ambient music. No text, no logo, no watermark.

2. Running shoe on the move

A neon-green running shoe hits a wet asphalt track at dawn, splashing a thin spray of water. Low-angle close-up at ground level, smooth lateral tracking shot alongside the shoe as the runner keeps a steady pace. Cool blue morning light with warm highlights, high-contrast sports commercial look. Sound: rhythmic footsteps and light breathing. Keep it subtitle-free.

3. Headphones unboxing

Two hands lift the lid of a matte black box, revealing over-ear headphones resting in grey foam. The hands slowly lift the headphones out and turn them toward the camera. Top-down shot, fixed camera position. Soft diffused light, dark charcoal desk, premium tech-ad style. Sound: the soft rustle of the box and a low synth tone. No logo, no watermark.

Food and drink

4. Ramen close-up

A steaming bowl of ramen on a dark wooden counter. Chopsticks lift a strand of noodles slowly, broth dripping back into the bowl. Close-up, slow push-in toward the noodles. Warm tungsten light from above, steam glowing in the backlight, shallow depth of field, cinematic food photography. Sound: soft kitchen ambience with a pan sizzling in the background.

5. Iced coffee pour

Cold milk pours into a tall glass of iced coffee, swirling into slow marbled clouds around the ice cubes. Macro shot at eye level, fixed camera position. Bright natural window light, cream and caramel tones, clean café commercial style. Sound: ice cubes clinking and the soft pour of milk.

Fashion and beauty

6. Linen dress in the wind

A woman in a loose white linen dress and a wide straw hat walks slowly along a sand dune. The wind gently lifts the hem of her dress; she raises one hand to hold her hat. Full shot, camera follows her from the side. Golden-hour backlight, warm desaturated tones, editorial fashion film style. Sound: soft wind and distant waves.

7. Lipstick swatch

Extreme close-up of a deep red lipstick being twisted up from a gold tube, then swatched in one smooth stroke on the back of a hand. Fixed camera position, macro detail. Soft beauty lighting, blush-pink background, glossy high-end cosmetics ad style. No text, no logo, no watermark.

Characters and dialogue

8. Café confession

A young man in a grey hoodie sits across from a woman in a mustard sweater at a small café table. He fidgets with his cup, then looks up and says, "I got the job. I'm moving to Lisbon." She freezes, then slowly smiles. Medium close-up, over-the-shoulder shot from behind the woman. Warm afternoon light through the window, natural colors, indie film style. Sound: quiet café chatter and the hiss of an espresso machine. Keep it subtitle-free.

9. Grandmother's recipe

An elderly woman with silver hair in a bun kneads dough on a flour-dusted table, her hands moving slowly and confidently. She glances at the camera and says warmly, "The secret is patience." Medium shot, slow push-in toward her face. Soft window light, warm earthy tones, documentary style. Sound: the soft thud of dough and a ticking kitchen clock.

10. Detective at the window

A detective in a long dark coat stands at a rain-streaked office window at night, holding a photograph. He lowers the photograph slightly, jaw tense, and says quietly, "She was here last night." Medium close-up from the side, fixed camera position. Neon light from the street in blue and red, deep shadows, film-noir style with film grain. Sound: steady rain and a distant siren.

Landscape and travel

11. Mountain lake at sunrise

A still alpine lake reflecting snow-capped peaks at sunrise. Thin mist drifts slowly across the water surface. Aerial drone shot gliding forward low over the lake. Pink and gold sunrise light, crisp air, nature documentary style, rich detail. Sound: soft wind and a single bird call.

12. Night market walk

A busy night market street with glowing lanterns, food stalls and steam rising from grills. People stroll past slowly. First-person POV walking forward at a calm pace through the crowd. Warm lantern light against a deep blue night sky, vibrant colors, travel vlog style. Sound: sizzling food, chatter and distant music.

Action and sports

13. Skateboard trick

A skateboarder in a red beanie rolls toward a concrete ledge in an empty skate park, crouches, and slides the board along the ledge before landing cleanly. Wide-angle lens, camera follows close behind at knee height. Late-afternoon sun, long shadows, gritty street-sports style. Sound: wheels on concrete and the clack of the board.

14. Boxer training

A boxer wraps her hands slowly in a dim gym, then throws three controlled punches at a heavy bag. Sweat glistens on her shoulders. Medium shot, handheld camera with slight realistic shake. A single overhead light, high contrast, desaturated tones, gritty sports-documentary style. Sound: punches hitting the bag and her steady breathing.

Anime and stylized

15. Anime rooftop

2D Japanese anime style. A high-school girl with short black hair and a navy uniform stands on a school rooftop at sunset, her scarf fluttering. She turns her head toward the camera and smiles faintly. Medium shot, camera slowly circles around her. Orange and purple sky, hand-painted background, soft cel shading. Sound: wind and distant cicadas.

16. Clay animation fox

Stop-motion clay animation style. A small orange clay fox tiptoes across a felt forest floor, sniffs a red mushroom, and jumps back in surprise. Full shot at the fox's eye level, fixed camera position. Warm miniature-set lighting, visible handmade textures, playful mood. Sound: soft footsteps and a cartoon "pop".

17. Cyberpunk alley

Cyberpunk style. A courier in a reflective jacket rides an electric bike slowly through a narrow alley full of holographic signs. Rain drips from cables overhead. Low-angle shot, camera tracks backward in front of the rider. Neon magenta and cyan lighting, wet reflective ground, cinematic sci-fi look. Sound: electric hum, rain and muffled city noise. Avoid generating any readable text or subtitles.

Multi-shot sequences

18. Morning routine (three shots)

Shot 1: Close-up of an alarm clock on a nightstand; a hand reaches in slowly and switches it off. Fixed camera position.
Shot 2: Cut to a medium shot of a man in a striped T-shirt opening the curtains; morning light floods the small apartment. Camera slowly pulls back.
Shot 3: Cut to a close-up of coffee pouring into a white mug on the kitchen counter. Slow push-in.
Warm natural light, soft colors, calm lifestyle-film style. Sound: the alarm beep, curtains sliding, coffee pouring, and gentle acoustic guitar. Keep it subtitle-free.

19. Lost dog found (three shots)

Shot 1: Wide shot of a girl in a yellow raincoat calling out in a rainy park, holding a red leash. Camera follows her from the side.
Shot 2: Cut to a low-angle close-up of a scruffy brown dog under a bench, ears perking up. Fixed camera position.
Shot 3: Cut to a medium shot as the dog runs to the girl and she kneels to hug it, laughing. Camera slowly rises above them.
Overcast soft light, muted greens and yellows, heartwarming short-film style. Sound: rain, her voice calling "Biscuit!", and a happy bark.

Image-to-video and references

20. First-frame product shot

Use Image 1 as the first frame. The perfume bottle stays centered while light from the left slowly sweeps across the glass and the liquid inside shimmers. Camera slowly pushes in. Keep the bottle shape, label area and background exactly as in Image 1. No text, no logo, no watermark.

21. Character from a reference image

Define the illustrated girl with a green scarf and round glasses in Image 1 as Mia. Mia walks slowly through an autumn park, picks up a red leaf and holds it up to the light. Medium shot, camera follows her from the front. Keep Mia's face, scarf and glasses consistent with Image 1. Soft afternoon light, warm storybook illustration style. Sound: crunching leaves and light piano.

22. Camera move from a reference video

Reference the camera movement in Video 1. A vintage red car drives along a coastal cliff road at sunset. Keep the car's color and shape from Image 1. Warm golden light, sea haze, classic road-movie style. Sound: engine hum and waves below.

For the reference prompts (20–22), BytePlus says to name uploaded assets by type and number — "Image 1", "Video 1", "Audio 1" — rather than by asset ID, and to define a character once ("Define … in Image 1 as Mia") and reuse that label.

Common Seedance 2.0 prompt mistakes (and how to fix them)

1. Overloading one prompt

Before:

A woman runs through a city, jumps over cars, fights three robots, rides a motorbike, flies a helicopter and lands on a rooftop at sunset in a cinematic epic style.

After:

A woman in a leather jacket sprints across a rooftop at sunset and leaps a narrow gap between two buildings. Camera follows her from the side. Warm orange backlight, cinematic action style. Sound: wind and her footsteps.

One clip, one main event. Split the rest into separate shots or generations.

2. Conflicting camera moves

Before: Camera pushes in, pulls back, pans left and orbits around the singer.

After: Camera slowly orbits around the singer.

The official guide asks for one camera movement per shot because combined moves make the image unstable. For more moves, use Shot 1 / Shot 2.

3. A vague subject

Before: A person walks in a room.

After: A tall man in a navy raincoat and grey scarf walks slowly into a dim, book-lined study.

Two or three stable details give the model something to keep consistent.

4. Abstract emotions

Before: She is very sad.

After: She lowers her head, her shoulders tremble slightly, and she grips the corner of her sleeve.

BytePlus recommends showing emotion through visible body details instead of adjectives.

5. Forcing exact timings

Before: 0–2s: he opens the door. 2–5s: he sits. 5–8s: he speaks.

After: Shot 1: He opens the door. Shot 2: Cut to him sitting down at the table. Shot 3: Close-up as he says, "We need to talk."

The official guide notes precise timings are unstable; order the events and let the model pace them.

6. Leaving the model to guess about text and logos

Before: a product prompt with no constraints, which may come back with random captions or a watermark.

After: add Keep it subtitle-free, no logo, no watermark. BytePlus also notes that subtitles are less likely in landscape formats than in portrait.

7. Mixing languages in dialogue

Before: She says, "Bonjour, how are you today?"

After: She says, "Hello, how are you today?"

Keep each line of dialogue in one language; the official guide warns against mixing.

Turn your idea into a Seedance prompt

You don't have to write every prompt from scratch:

  • Have a rough idea? Paste it into our seedance prompt generator to get a structured Seedance prompt with subject, motion, camera and sound already separated.
  • Found a clip you want to recreate? Use video to prompt to reverse-engineer a prompt from an existing video, then adapt it with the formula above.
  • Writing for several models? Try the AI video prompt generator.

Seedance 2.0 Prompt FAQ

How do I use Seedance 2.0?

Open Seedance 2.0 on Dreamina, in CapCut's AI video tools, on Jimeng AI or Doubao in China, or call it through the BytePlus ModelArk or Volcengine Ark API. Pick text-to-video or upload reference images, clips or audio, write a prompt in the order subject, action, scene, light, camera, style and constraints, choose a length (4–15 seconds) and aspect ratio, then generate and refine one detail at a time.

Where can I access Seedance 2.0?

ByteDance describes Dreamina as the official website for Seedance 2.0. CapCut rolled it into its AI video features and expanded it to Japan on March 31, 2026 and to the US on April 7, 2026. In China it is available on Jimeng AI and Doubao, and developers can use it through BytePlus ModelArk or Volcengine Ark. Availability and plans vary by platform and region.

What is the difference between Seedance 2.0 and Seedance 2.5?

Seedance 2.5 launched on July 31, 2026. ByteDance Seed lists up to 30 seconds per generation (with extensions), up to 30 reference images, 10 video clips and 10 audio clips, and timestamp-level editing. Seedance 2.0 generates 4–15 seconds and accepts up to 9 images, 3 video clips and 3 audio clips. The prompt structure in this guide works for both; for 2.5-specific examples see our seedance prompt generator.

How long can a Seedance 2.0 video be?

BytePlus documents a duration of 4 to 15 seconds per generation for all Seedance 2.0 models, or automatic duration selection. Longer pieces are made by extending a clip or by generating separate shots and editing them together.

Does Seedance 2.0 generate sound?

Yes. Seedance 2.0 generates audio together with the video, including voices, sound effects and background music. Put spoken lines in quotation marks and describe the sound for each shot.

Can I use a photo of a real person as a reference?

BytePlus states that Seedance 2.0 models do not support direct uploads of reference images or videos containing real human faces, and CapCut restricts videos made from images or videos of real faces. Use illustrated characters, products, places or generated characters as references.

How long should a Seedance 2.0 prompt be?

BytePlus recommends no more than 1,000 English words (500 Chinese characters). Shorter is easier to refine, so write one clear shot and change one detail per generation.

Does Seedance 2.0 support negative prompts?

There is no separate negative prompt field in the BytePlus API. Write constraints in the prompt itself, for example "keep it subtitle-free", "do not generate a logo" or "do not generate a watermark".

Sources

  1. ByteDance Seed — Seedance 2.0 official launch (Feb 12, 2026)
  2. BytePlus ModelArk — Dreamina Seedance 2.0 series prompt guide
  3. BytePlus ModelArk — Seedance 2.0 model documentation
  4. BytePlus ModelArk — Video generation tutorial
  5. BytePlus ModelArk — Create video generation task (API reference)
  6. BytePlus ModelArk — Seedance lens language examples
  7. ByteDance Seed — Introducing Seedance 2.5 (Jul 31, 2026)
  8. Dreamina — Seedance 2.0
  9. CapCut Newsroom — Dreamina Seedance 2.0 in CapCut

More guides