Guide · prompting

Seedance 2.5 prompt guide, with 12 examples

Seedance 2.5 follows a prompt written like a shot list: who, what they do, where, how the camera moves, and what you hear. Because one generation can run up to 30 seconds, structure matters more than adjectives. Here is a template, the words that steer it, and 12 examples to copy.

The template

Number the shots and give each a framing, a camera move and an action. Put timings on long clips. End with the sound.

Prompt template
[Style or look, if not realistic]. [Subject] [does action] in [place], [time of day / light].
Shot 1 (0–5 s): [framing], [camera move]. [What happens].
Shot 2 (5–10 s): [framing], [camera move]. [What happens].
Shot 3 (10–15 s): [framing]. [What happens]. [Line of dialogue in quotes, if any].
Sound: [what runs under the whole clip], [sounds at specific moments].

What each part does

Seedance 2.5 prompt parts
PartWriteExample
SubjectWho or what, with one or two defining detailsa barista with a shaved head and a green apron
ActionOne clear action per shot, in the present tensepours steamed milk into a cup, drawing a leaf
Place and lightWhere, and the light sourcein a small cafe, morning sun through the window
CameraFraming plus one moveclose-up, slow push-in
ShotsNumbered, with seconds on long clipsShot 2 (5–10 s): ...
StyleOnly if not photoreal2D anime, film grain, claymation, 3D motion graphics
SoundA bed under the clip plus momentssoft jazz, cups clinking, a door bell at the end
DialogueThe exact words in quotes, who says themShe says: "We open at seven."

Camera words that work

Framing: extreme close-up, close-up, medium shot, wide shot, aerial, low angle, overhead, over-the-shoulder, POV. Moves: static / locked-off, slow push-in, pull back, pan left or right, tilt up, orbit, tracking shot, dolly, crane up, handheld, gimbal walkthrough, whip pan. Speed: slow motion, time-lapse, real time. Pick one move per shot; two moves in one shot tend to fight.

Sound

Seedance 2.5 generates sound with the picture when audio is true, at no extra charge. Describe three layers: the bed (music genre, room tone), the moments (a door slam on shot 3), and any dialogue in quotes with who says it. Short lines work best; one or two sentences per speaker in a 10 second clip.

Long clips (15 to 30 seconds)

  • Give every shot a time range so the pacing is yours, not the model’s.
  • Three to five shots for 30 seconds. More cuts than that and each one gets too short to read.
  • Keep one or two characters, and restate a key detail (the red coat) in each shot they appear in.
  • Pin faces and products with reference images rather than long descriptions.
  • Draft the opening shots as a 10 second 480p clip ($0.90) before rendering 30 seconds at 1080p ($15.60).

What to avoid

  • Text in the frame. Signs, captions, prices and logos come out garbled; add them in your editor.
  • Lists of adjectives. “Cinematic, epic, 8k, masterpiece” adds little. Describe the light and the lens instead.
  • Contradictions. “Static shot” and “the camera flies through the city” in one shot.
  • Crowds that matter. Ten named characters will not stay consistent; keep the important ones to two.
  • Negations. “No people” can put people in. Say what is there: “an empty street”.

12 Seedance 2.5 prompt examples

Paste any of these as input.prompt with audio: true. Set the duration to the length in the title.

Product ad (15 s, 9:16)

Shot 1: macro of water droplets running down a cold glass bottle on wet black stone. Shot 2: the bottle spins as a hand grabs it, slow motion. Shot 3: a woman on a sunny rooftop drinks and laughs, handheld. Ice cracking, a fizz, upbeat electronic beat building to a drop on shot 3.

UGC-style talking head (10 s)

A woman in her late twenties films herself on a phone in a bright kitchen, holding a jar of face cream up to the camera. She says: "Okay, I did not expect this to work this fast." Natural daylight, slight handheld shake, casual tone, room echo.

Real estate walkthrough (20 s, 16:9)

Smooth gimbal walkthrough of a modern apartment. Shot 1: the front door opens onto a bright living room with floor-to-ceiling windows. Shot 2: glide past the kitchen island. Shot 3: step onto the balcony, reveal the city skyline at golden hour. Soft piano, quiet street sounds below.

Food close-up (8 s)

Close-up of a chef pouring thick chocolate sauce over a stack of pancakes, slow motion, steam rising, shallow depth of field, warm tungsten light. Sauce drips, a fork clinks on the plate.

Anime fight (12 s)

2D anime style. Two swordsmen face each other on a rooftop at night in the rain. Shot 1: wide, lightning flashes. Shot 2: they dash, blades clash in a burst of sparks, speed lines. Shot 3: close-up of one smiling as his hat falls. Rain, thunder, steel ringing.

Music video (30 s)

Shot 1 (0–8 s): a singer in a red coat walks through an empty neon-lit subway car, camera tracking backward. Shot 2 (8–16 s): she sits, the lights flicker in time with the beat. Shot 3 (16–24 s): the train exits into daylight over a bridge. Shot 4 (24–30 s): she looks into the lens. Synth-pop track, train rumble under the music.

Explainer (10 s)

Clean 3D motion graphics on a white background: a coin drops into a glass piggy bank, the bank fills with glowing coins, a line chart rises behind it. Friendly male voice: "Small amounts, every week, add up." Soft clicks, a light marimba.

Game trailer (15 s)

Dark fantasy. Shot 1: a knight walks through a ruined cathedral, dust in shafts of light. Shot 2: a dragon’s eye opens in the shadows. Shot 3: the knight raises a glowing sword as the dragon roars, camera shakes. Choir, deep drums, the roar fills the frame.

Fashion lookbook (12 s, 9:16)

A model walks toward the camera on a windy beach in a flowing white linen suit, slow motion, overcast soft light, film grain. Shot 2: close-up of the fabric moving. Waves, wind, a minimal house beat.

Nature documentary (20 s)

Shot 1: aerial over a misty rainforest at dawn. Shot 2: a red-eyed tree frog on a leaf, macro, a drop of water falls. Shot 3: the frog leaps out of frame. Birds, insects, dripping water, a calm narrator: "Every morning begins with a choice."

Horror short (15 s)

Found-footage style, night vision. A hallway in an abandoned hospital. Shot 1: slow walk forward, flashlight beam shaking. Shot 2: a wheelchair at the far end slowly rolls toward camera on its own. Shot 3: the light cuts out. Heavy breathing, a squeaking wheel, silence.

Sports highlight (10 s)

A basketball player drives to the hoop in an empty gym, low angle, slow motion as he leaves the ground, then normal speed for the dunk, the rim shaking. Sneakers squeak, the ball bounces, a single echoing clap.

Run a prompt

cURL
# One Idempotency-Key per job, made once: resend with the same $KEY after a timeout, a 429 or a 5xx
KEY=$(uuidgen)
curl https://api.revana.si/v1/generations \
  -H "Authorization: Bearer $REVANA_API_KEY" \
  -H "Content-Type: application/json" \
  -H "Idempotency-Key: $KEY" \
  -d '{
  "model": "seedance-2.5",
  "input": {
    "prompt": "Close-up of a chef pouring thick chocolate sauce over a stack of pancakes, slow motion, steam rising, shallow depth of field, warm tungsten light. Sauce drips, a fork clinks on the plate.",
    "duration": 8,
    "resolution": "480p",
    "audio": true
  },
  "max_cost_usd": 1
}'

That 8 second 480p draft costs $0.72 ($0.09 a second). From Claude or ChatGPT, paste the prompt and ask the Revana MCP server to quote and make it.

FAQ

How do I write a good Seedance 2.5 prompt?

Write it as a shot list: number the shots, give each a framing, one camera move and one action, add timings for clips over 10 seconds, and finish with the sound and any dialogue in quotes.

How long should a Seedance 2.5 prompt be?

Long enough to describe each shot, no longer. A 10 second clip usually needs 40 to 80 words; a 30 second, four-shot clip 100 to 180. Extra adjectives add little.

Can Seedance 2.5 do several shots in one video?

Yes. Numbered shots in one prompt become cuts inside one clip of up to 30 seconds.

Can Seedance 2.5 make characters speak?

Yes, with audio on. Put the words in quotes and say who speaks them; keep lines short.

How do I keep a character the same across shots?

Send reference images of the character (several angles) in image_urls, name them in the prompt ("the man in image 1"), and restate a defining detail in each shot.

Make your first Seedance 2.5 video

One API key for Seedance 2.5 and every other video and image model. Prepaid credit from $10, free signup credit, no subscription, and only delivered clips are charged.