Seedance 2.0 Prompts: 15 That Still Work in the 2.5 Era
Seedance 2.0 prompts still work on 2.5 if you fill five slots: subject, action, camera, style, audio. Get the formula, 15 original prompts, and what changed.
People still search for Seedance 2.0 prompts, even though the model chip in Dreamina now reads Dreamina Seedance 2.5. That's less strange than it sounds: a good video prompt is a structure, not a magic phrase, and the structure survived the version bump.
Seedance is ByteDance's video-generation model family, the engine behind Dreamina — the international version of Jimeng (即梦). Seedance 2.0 was the previous flagship generation; our Jimeng AI prompts guide covers 2.5, the one serving today.
Below: a five-slot formula — subject, action, camera, style, audio/atmosphere — and 15 original prompts written to it, each with a note on why it works. Treat them as strong first drafts, not guaranteed outputs.
The five-slot prompt formula
Every prompt below fills five slots, in this order:
| Slot | What it answers | Example | Common miss |
|---|---|---|---|
| Subject | Who or what, with one or two specifics | a matte-black espresso machine on a walnut counter | "A product" — nothing to light or frame |
| Action | What moves, and how fast | a single shot pours as steam curls upward | No verb, so you get a moving photo |
| Camera | Shot size, angle, one move | slow push-in to a close-up | Three moves in one short clip |
| Style | Look, light, lens, grade | shallow depth of field, warm grade | "Cinematic" as the whole slot |
| Audio / atmosphere | What we hear, how it feels | café hum, the hiss of the steam wand | Leaving the mood to chance |
Three rules keep the slots honest:
- One camera move per shot. Need two moves? You need two shots.
- Concrete nouns beat adjectives. "Sodium streetlight" gives the model something to render; "beautiful lighting" doesn't.
- English, 40 to 80 words. English proved most reliable in our Dreamina walkthrough, and longer prompts tend to contradict themselves.
15 Seedance 2.0 prompts by use case
Copy as written, then change one slot at a time.
Product shots
1. Espresso machine hero
A matte-black espresso machine on a walnut counter in morning light. A single shot pours into a glass cup as steam curls upward. Slow push-in to a close-up on the crema. Clean commercial look, shallow depth of field, warm grade. Audio: café hum, the hiss of the steam wand.
One object, one motion, one camera move — and the push-in ends on the detail that sells it.
2. Earbuds on a turntable
White wireless earbuds in an open charging case on a matte-gray turntable. The turntable makes one slow half turn, light sliding across the glossy case, then the lid eases shut. Locked-off macro shot at eye level. Minimalist studio look, soft overhead key light. Audio: quiet room tone, one magnetic click.
The product moves, so the camera doesn't. The single click gives the clip an ending.
3. Serum bottle in water
A frosted glass serum bottle stands in shallow water on a pale stone slab. A droplet falls and rings spread around its base. Low-angle close-up, slow slide from left to right. Clean spa aesthetic, diffused daylight, soft mint and white palette. Audio: one clear water drop, calm ambient tone.
The liquid does the moving. A named palette does more for brand consistency than "luxurious" ever will.
Cinematic scenes
4. Rain-soaked alley
A woman in a tan trench coat walks away from camera down a narrow alley at night, rain falling through the glow of one sodium streetlight. She pauses and glances back. Slow dolly forward at shoulder height. Neo-noir look, wet pavement, teal and amber grade. Audio: steady rain, distant traffic, echoing footsteps.
Two beats of action give the clip a shape; one named light source gives the grade an anchor.
5. Lighthouse keeper
An elderly lighthouse keeper climbs a spiral iron staircase holding a brass lantern, then looks out at a storm through salt-streaked glass. Low-angle shot from the bottom of the stairwell, tilting up as he rises. Painterly realism, cold blue storm light against warm lantern glow. Audio: howling wind, crashing waves.
The warm-cold contrast is written in, and the tilt follows his direction of travel.
6. Night train
A young man sits by a night-train window, city lights streaking past behind his faint reflection. He slowly folds a handwritten letter and closes his eyes. Static medium close-up from the seat opposite. Indie-film look, 35mm grain, muted greens, soft orange highlights. Audio: rail clatter, a muffled station announcement.
Motion lives in the background and one small gesture, so a static camera still feels alive.
Food close-ups
7. Ramen noodle lift
A steaming bowl of tonkotsu ramen on a dark wooden counter. Chopsticks lift glistening noodles from the broth as steam drifts across frame. Eye-level macro close-up, slow push-in as the noodles rise. Moody food photography, one warm side light, deep shadows. Audio: simmering broth, the murmur of a busy ramen shop.
Food sells through texture and steam, and side light is what makes steam visible.
8. Molten chocolate cake
A chocolate lava cake on a white plate, dusted with powdered sugar. A spoon breaks the crust and molten chocolate flows slowly out. Tight close-up at a 45-degree angle, camera completely still. Glossy dessert-commercial look, bright soft light, rich brown tones. Audio: a soft crack, then warm background music.
The flow is the payoff, so the camera stays still; "slowly" sets how long it lasts.
Travel and drone
9. Coastal cliffs reveal
Aerial drone shot over turquoise water and white limestone cliffs, a small sailboat cutting a wake below. The drone glides forward and slowly rises to reveal a whitewashed village on the clifftop. Smooth forward-and-up move, no rotation. Bright travel-documentary look, natural colors, midday sun. Audio: wind, distant waves, faint seagulls.
Drone prompts fail when the camera has no destination. "Forward and up" plus a reveal target gives it one.
10. Night market walk-through
A crowded night market with lantern-lit food stalls, vendors flipping skewers over open flames. A traveler in a denim jacket weaves through the crowd holding a cup of bubble tea. Handheld follow shot from behind, gentle sway. Vibrant street-photography look, warm tungsten light, light haze. Audio: sizzling grills, vendors calling out.
Shooting from behind sidesteps face consistency, and "gentle sway" reads as vlog rather than stabilizer.
Character action
11. Skateboard kickflip
A teenage skateboarder in a red hoodie rolls across an empty concrete skatepark at golden hour and lands a clean kickflip. Low-angle tracking shot moving alongside at board height. Gritty 16mm skate-video look, warm sun flare, light film grain. Audio: wheels rumbling on concrete, the sharp clack of the landing.
Fast action is the riskiest category, so name one trick and pin the camera beside it.
12. Kata in a bamboo forest
A martial artist in a white uniform performs a slow, controlled kata in a misty bamboo forest at dawn, ending in a still, balanced stance. Wide shot slowly arcing a quarter circle around her. Serene look, soft diffused light, pale green tones. Audio: rustling bamboo, birdsong, the swish of fabric.
Slow moves in a wide frame forgive a lot, and the final pose gives the clip a last frame.
Documentary style
13. Potter at the wheel
An elderly potter with clay-covered hands shapes a tall vase on a spinning wheel. She smooths the rim with a sponge, then leans back to check the shape. Subtle handheld medium shot, as if from a documentary crew. Observational style, dusty window light, earthy tones. Audio: the whir of the wheel, a faint radio.
Documentary realism is restraint: subtle handheld, available light, and an ordinary second action an ad would cut.
Stylized and anime
14. Anime rooftop at sunset
Anime style: a student with short blue hair stands on a school rooftop at sunset as cherry petals swirl past in a sudden gust. She turns to the camera with a small smile. Slow push-in from wide to medium close-up. Hand-drawn 2D look, soft cel shading, saturated orange sky. Audio: wind, a distant school bell.
Lead with the style tag, then restate it with specifics so the look doesn't drift toward 3D.
15. Claymation fox
Stop-motion claymation: a small clay fox in a knitted scarf pushes a wheelbarrow of pumpkins across a felt autumn hill, then stops to wipe its brow. Static wide shot, locked like a tabletop setup. Handmade look, visible fingerprints in the clay, slightly choppy frame rate. Audio: plucked ukulele, a squeaky wheel.
Stop-motion is defined by its imperfections. Ask for them outright, or you get smooth 3D wearing a clay texture.
From 2.0 to 2.5: what changed and what didn't
Seedance 2.0 is ByteDance's earlier-generation video model, in the same family as the 2.5 that Dreamina serves now. We're not printing a 2.0 spec sheet: numbers for superseded versions mostly circulate secondhand, and we only publish what we can verify.
Treat the jump as a generation step, not a new language. Seedance 2.0 prompts carry straight across; nothing in the prompts above depends on a version-specific feature. What the newer generation rewards is precision in the camera and audio slots: "slow dolly forward at shoulder height" and "steady rain, distant traffic" pay off where "cinematic" and "nice music" never did.
The bigger change is the workflow around the prompt:
- References. Up to 50 — images, videos and audio — pinned into the text with @ mentions.
- Label weaving. Codename each asset in caps (JESSICA, CREDIT_MODULE) and attach files with matching names. The subject slot becomes a codename; the other four stay as written.
- Long video mode. Credits scale with duration — our walkthrough's 60-second 720P clip listed at around 4,000 credits at recording time. Validate at 480P and short durations first.
- Timestamps are optional. For montage, the version with every timestamp stripped came out cleanest in our guide's three-round test.
Where to go next
For the full current-generation workflow, read our Jimeng AI prompts guide for Seedance 2.5. The five slots also transfer to other video models: see our Kling 4 prompts guide and Sora 2 prompting guide, or compare Runway and Kling AI directly.
FAQ
Is Seedance 2.0 still available in Dreamina?
Dreamina's composer defaults to Dreamina Seedance 2.5, and we haven't verified whether 2.0 is still selectable anywhere. Plan around 2.5 — Seedance 2.0 prompts carry over because the structure is the same.
Should Seedance prompts be written in English?
Yes, as a default. Our Dreamina walkthrough found English the most reliable, with complex Chinese sentences occasionally parsed wrong. Keep the prompt body English-first.
Why does the same prompt give different results?
Video generation isn't deterministic, so no model renders a prompt identically twice. Change one slot per run, and test at low resolution before paying for a long render.
What's the difference between Jimeng, Dreamina and Seedance?
Seedance is the model family. Jimeng (即梦) is ByteDance's app for the Chinese market, and Dreamina is its international version — one product, two names.
