Seedance 2.0 Prompt Examples
Sixteen copy-paste prompts sorted by content type, built on how Seedance 2.0 actually reads a prompt — action first, energy named, one visual anchor.
Seedance 2.0 rewards a different kind of prompt than the heavyweight cinematic models. Describe the action plainly, name the energy, anchor the subject with one distinctive visual, and let the model pick the camera — that is most of the trick. This guide collects sixteen copy-paste Seedance 2.0 prompt examples sorted by the kind of content they make, each with a short note on why it works, plus the settings and iteration habits that turn the model's speed into an actual advantage.
Every prompt below runs today on ByteDance's speed-tier video model. Paste one in, swap the subject for yours, and you will have a first take in well under a minute.
How Seedance 2.0 reads a prompt
Seedance 2.0 was optimized for the kind of motion that performs on short-form platforms, and its prompting style follows from that. Where some models reward elaborate cinematographic direction, Seedance works best when you describe what happens rather than how to film it — give it a clear action and it infers sensible camera work on its own. A single camera cue is fine; a full shot list is wasted effort here.
Three other habits pay off consistently. Energy words visibly change the output: "energetic," "calm," "playful," and "dramatic" shift the pacing and motion quality, not just the mood. Simple scenes beat crowded ones, because complex multi-object interactions are where artifacts creep in. And strong visual anchors — a bright red leather jacket, a mint green bicycle — give the model something concrete to hold onto, so subjects stay clean across the clip.
| Layer | What to write | Example |
|---|---|---|
| Subject with an anchor | One subject, one distinctive visual | "a dancer in a bright red leather jacket" |
| Action | The single thing that happens | "spins into a freeze under a streetlight" |
| Energy | One or two mood words | "high-energy, playful" |
| Setting and light | Where it happens, what the light does | "wet asphalt reflecting neon at night" |
| Format | Aspect ratio for the destination | "9:16" |
Write those layers as one flowing sentence rather than a checklist — Seedance reads natural language well, and 40 to 60 words is the sweet spot.
Vertical social prompts (TikTok, Reels, Shorts)
Vertical is where Seedance 2.0 is most at home. The model was trained with 9:16 as a first-class format, so its vertical-native mode frames people naturally and leaves room for text overlays and platform UI instead of cropping a landscape composition.
A barista in a denim apron slides a caramel latte across the counter and grins at the camera, playful energy, cozy cafe with warm string lights, morning sun through the window. 9:16.
One subject, one action, one light source. The grin at the camera gives the clip a natural hook frame for a caption.
A woman in a yellow raincoat jumps over puddles on a rainy city sidewalk, laughing, carefree energy, soft overcast light, storefronts glowing in the background. 9:16.
The raincoat is the anchor; the overcast light keeps skin tones stable while the puddle jump supplies the motion beat.
A skateboarder in a tie-dye hoodie lands a kickflip on a sunlit basketball court and throws both arms up, triumphant energy, golden afternoon light, long shadows. 9:16.
Ending the action on a pose gives the clip a clean out-frame — useful when you loop or cut to a product card.
A college student unboxes a mystery package at a cluttered desk, eyes widening in delight, excited energy, ring-light glow, posters on the wall behind. 9:16.
Reaction beats are a Seedance strength: facial expressions read genuinely, and the unboxing format maps straight onto affiliate and review content.
Dance and movement prompts
Full-body movement is the model's standout lane — ByteDance tuned it on exactly this kind of content, and hand gestures, rhythm, and weight all render above its class.
A breakdancer in a silver puffer jacket hits a windmill in the middle of a neon-lit parking garage, high-energy, headlights streaking past in the background. 9:16.
Name the specific move. "Hits a windmill" produces committed rotation where "dances impressively" produces vague shuffling.
Three friends in matching white sneakers do a synchronized shuffle dance on a rooftop at sunset, joyful energy, city skyline behind them, warm backlight. 9:16.
Matching wardrobe is doing double duty here: it is a visual anchor and it hides the small character drift that group shots can show.
A ballerina in a deep green dress turns a slow pirouette in an empty marble hall, calm and dramatic, shafts of window light cutting through dust. 16:9.
Slowing the energy down works too — "calm and dramatic" pulls the pacing back and lets the light do the composition.
Product prompts
Product clips want the opposite of dance clips: one object, one clean surface, and light described explicitly, because there is no human action for the model to infer a camera from.
A matte black wireless earbud case pops open on a white marble table, one earbud rising slightly as if lifted, clean minimal energy, bright softbox lighting. 1:1.
A single mechanical action — the case popping open — is enough motion. Square format drops straight into a product grid.
A glass bottle of cold brew tips forward and pours over ice in slow motion, coffee swirling through the cubes, rich indulgent mood, dark background with a single warm spotlight. 9:16.
Liquids are forgiving on Seedance when they are the whole subject; keep the background dark so the pour owns the frame.
A running shoe in electric orange rotates on a concrete pedestal as dust particles drift through a beam of light, bold energetic mood, dramatic side lighting. 1:1.
Rotation plus drifting particles is a reliable recipe: steady primary motion with a secondary layer that adds depth without risking artifacts.
Lifestyle and b-roll prompts
These are the filler shots that make an edit feel produced — cheap to generate here, expensive to film.
A golden retriever shakes off water in a sunlit backyard, droplets flying, joyful energy, green lawn and warm evening light. 16:9.
Steam rises from a bowl of fresh ramen as chopsticks lift the noodles, cozy appetizing mood, soft window light in a small kitchen. 9:16.
Rain streaks down a cafe window while a candle flickers on the sill inside, calm reflective mood, blurred umbrellas passing outside. 16:9.
A cyclist in a mint green jacket rides through falling autumn leaves on an empty park path, serene energy, low golden light between the trees. 16:9.
The pattern across all four: one continuous motion (shaking, rising steam, streaking rain, riding) rather than a sequence of events. Eight seconds is one beat, not a scene.
Image-to-video starter prompts
When the look matters more than anything — a real product, a brand character — lock it by starting from a still frame and let the prompt describe only the motion.
The model in the photo turns her head toward the camera and smiles as a breeze lifts her hair, gentle energy, keep the outfit and lighting exactly as the reference.
The product in the image slowly rotates a quarter turn while the background lights pulse softly, keep colors and framing from the reference.
Notice both prompts spend their words on motion and explicitly pin everything else to the reference. That division of labor — image owns the look, prompt owns the movement — is the habit that scales when future Seedance versions accept more references.
Settings that matter
| Setting | Seedance 2.0 | Practical note |
|---|---|---|
| Resolution | Up to 720p | 1280x720 landscape, 720x1280 portrait |
| Aspect ratios | 16:9, 9:16, 1:1 | Set it in the prompt or the generator |
| Clip length | Up to 8 seconds | One beat per clip; chain clips for scenes |
| Frame rate | 24fps | Standard playback everywhere |
| Audio | None | Add music and voiceover in your edit |
Two of these shape your workflow. The 720p ceiling is a non-issue for phone-first platforms, but if a clip is headed for a large screen, run the keeper through a video upscale pass rather than generating everything at a heavier model's cost. And since there is no native audio, plan the sound in your editor — most social creators overlay their own tracks anyway, so in practice this costs less than it reads.
The iteration loop that makes Seedance worth it
Seedance 2.0's real advantage is not any single output; it is renders that finish inside a minute, which changes how many ideas you can afford to test. Ten prompt variations in ten minutes is a realistic session, and that volume beats careful prompt-polishing on a slow model almost every time.
The loop that works: write the prompt from the recipe above, generate, and judge only the base motion on the first take. If the action reads, refine one layer at a time — swap the energy word, or the light, or the anchor, never all three at once, so you can tell which change did the work. For bulk drafting, the draft-speed Fast variant stretches a credit balance further; re-run only the winning prompt on the full model. When you want to try the whole loop now, run any of these prompts with Seedance preloaded.
Common Seedance 2.0 prompt mistakes
- Directing the camera instead of the action. A stacked shot list ("dolly in, then crane up, then orbit") fights the model. Describe the action; add at most one camera cue.
- Crowding the scene. Four subjects interacting is artifact territory. One subject doing one thing, cleanly, wins.
- Adjective soup without an anchor. "Beautiful, cinematic, stunning" gives the model nothing to render. One concrete visual detail outworks five style words.
- Ignoring the destination format. A landscape clip cropped to vertical loses the framing Seedance would have gotten right natively. Set 9:16 up front for social.
Will these prompts work on Seedance 2.5?
Yes — structurally, nothing changes. Seedance 2.5 was announced on June 23, 2026 and remains in closed beta with no public access as of late July 2026, so every prompt on this page targets the live 2.0 model. The reveal features — single-pass 30-second clips, up to 50 reference inputs, region-level editing — expand scale rather than syntax, and what ByteDance actually announced for 2.5 suggests reference habits will matter even more. For the version-agnostic prompt anatomy, the 2.5-focused prompt structure guide is the companion to this page: same recipe, fewer examples, more theory. Learn the loop on 2.0 now and the upgrade lands on top of a working routine.
