Skip to main content
Seedance 2.0

AI Video
Generation for Creators & Pros

Seedance 2.0 renders clips where audio syncs to the frame, characters stay consistent across cuts, and the controls match what a production workflow actually demands.

Capabilities

Contextual inline references and multimodal input within a single prompt

Inline Reference Prompting

Key Features of Seedance 2.0

Multimodal prompting, native audio sync, multi-shot coherence, narrative control

Multimodal Input & Control

Use text, images, audio, and video together as creative prompts. Provide storyboards, voiceovers, and motion samples inline — Seedance 2.0 honors their context and order in the final clip. This isn't simple reference attachment; it's interpretive multimodal prompting.

Native Audio-Visual Synchronization

Audio and video aren't stitched afterward — they're generated together. That leads to real lip sync, natural soundscapes, and dramatic timing built into the model's core workflow. Every frame and every sound share the same generative origin.

Multi-Shot Consistency

Seedance 2.0 keeps character details, lighting, style, and motion consistent across all generated shots — a key improvement over older models that suffered from "morphing" artifacts between scenes. Same face, same outfit, same world across every cut.

Narrative Understanding & Control

The model interprets prompts with structural intent — story logic, camera motion, cuts, and pacing are reflected in the output. Describe a sequence, and Seedance 2.0 builds it with cinematic awareness, not just raw visual generation.

Practical Output for Professionals

Generate clips up to approximately 15 seconds with real production value — perfect for ads, creative teasers, social content, and previsualization. The output is designed to be usable, not just impressive as a demo.

Compare Models

Seedance 2.0Kling 3.0Veo 3
DeveloperByteDanceKuaishouGoogle DeepMind
Audio-Video SyncUnified generationBuilt-in generationBasic
Multi-Shot Coherence
Multimodal InputText + Image + Audio + VideoText + Image + VideoText + Basic
Narrative Control
Max Duration~15 sec~15 sec~8 sec

Seedance 2.0 prompt examples

View all 23
Aug 31, 202630
Read the full promptBlue Dial Watch Through SmokeLuxury product reveal. A bold white sans-serif headline stays burned across the top of the frame for the whole shot and reads exactly: Make premium product reveals. Below it, a generic unbranded wristwatch with a polished steel case, a deep blue sunburst dial and a steel bracelet rests on a dark stone pedestal in a black studio. The face of the dial is a completely bare, smooth, empty surface of deep blue: apart from the plain polished baton hour markers and the two hands there is absolutely nothing printed on it — no brand name, no maker's name, no logo, no emblem, no crown, no word, no letter, no digit, no numeral, no date window, no small print at the top of the dial, no small print at the bottom of the dial, and no writing of any kind, not even faint or blurred lettering. The same is true in every macro close-up: when the camera comes close to the dial the surface is still bare blue with no text on it at all. Nothing is engraved or printed on the case, the crown, the bracelet or the clasp either. Do not depict any real brand, trademark or logo anywhere in the frame. Thick smoke rolls low across the floor and drifts. A single hard beam of warm light cuts through the haze and finds the watch, lifting it out of the darkness. The camera orbits slowly around the pedestal; a crisp rim light slides along the case and the bracelet, sharp macro glints flare on the crystal and the crown. The smoke parts, the light widens, and the watch settles into a clean centered hero shot, whole and complete with clear space around it, and the camera holds still on it for the last full second. Cinematic slow motion, rich blacks, glossy reflections, premium commercial cinematography, shallow depth of field. No people, no dialogue, no subtitles. The headline at the top is the only text in the entire frame.
Aug 31, 202610
Read the full promptCobalt Coat Campaign In 3 LooksA bold white sans-serif headline stays burned across the top of the frame for the whole shot. The headline is spelled letter by letter as C-A-M-P-A-I-G-N and reads exactly: Generate fashion campaign videos. Below it, a high-fashion campaign film, three bold looks cut back to back. One: a female model in an oversized electric cobalt-blue tailored coat with huge sculptural shoulders, worn over a liquid silver slip dress and tall glossy boots, strides straight toward the camera down a sunlit city street while the coat flares open behind her in the wind; the camera tracks backwards ahead of her, low and wide. Hard cut: the same model in a sculptural acid-yellow leather trench with an exaggerated collar, standing in a dark studio lit by hard coloured gels — a magenta rim light on one side, a cyan wash on the other — she turns her shoulder sharply and the fabric snaps. Hard cut: she sweeps through a long glass corridor in a floor-length emerald green sequinned gown with a fringed hem that shivers with every step, the sequins throwing points of light across the walls. Strong graphic styling: sharp eyeliner, slicked hair, statement silhouettes, wind machine on every shot. Saturated editorial colour grade, deep contrast, film grain, glossy magazine cinematography, confident runway attitude. She is an ordinary anonymous person and looks like nobody famous. Do not depict any real person, actor, celebrity or public figure, and do not reproduce anyone's likeness. All clothing is plain and unbranded, no logos, no monograms, no shop signs and no lettering in the background. Do not depict any real brand or trademark. No dialogue, no subtitles. The headline at the top is the only text in the entire frame.
Aug 31, 202630
Read the full promptCream Melting Into The HandBeauty and skincare commercial. A bold white sans-serif headline stays burned across the top of the frame for the whole shot and reads exactly: Create beauty and skincare videos. Below it, an extreme close-up of glowing skin on a cheekbone, soft diffused daylight grazing across it, fine dewy texture and tiny highlights on the surface. A single drop of clear serum slides slowly down the skin in slow motion. Cut to a plain unbranded frosted glass dropper bottle and a small ceramic jar of white cream on a pale stone surface, soft shadows, a few green leaves beside them. The bottles and jars are completely blank: no brand name, no label, no logo, no lettering anywhere. Do not depict any real brand, trademark or logo in the frame. The index fingertip of the right hand dips into the open jar and lifts a small soft swirl of white cream. Cut to both hands in soft daylight above the pale stone surface. The left hand rests palm down, relaxed, with the back of it facing up towards the camera; the right hand comes in from the other side and its index fingertip carries the cream onto the back of that left hand, spreading it there in one slow, gentle circular stroke. The two hands are a correct natural pair — one right hand and one left hand, mirror images of each other, with the thumbs on opposite sides. They are never two right hands and never two left hands. Each hand has exactly five fingers, normal human proportions and correct anatomy, and both hands clearly belong to the same person. The white cream thins out into a fine sheen, melts in and disappears, leaving a soft dewy glow on the back of the hand. The stroke slows down and comes to rest, the right hand lifts away, and the shot settles for a moment on the calm finished left hand. The ending is a complete, settled image — the movement finishes on screen and nothing is cut off halfway through a gesture. Only hands and a small patch of glowing skin are ever visible. Do not depict any real person, actor, celebrity or public figure, and do not reproduce anyone's likeness. The camera drifts in slowly, shallow depth of field, warm neutral palette of cream, beige and soft pink, clean minimal beauty cinematography, high-end commercial look. No dialogue, no subtitles. The headline at the top is the only text in the entire frame.
Aug 31, 202640
Read the full promptHoodie To Gown In One SpinSocial media transformation video. A young woman in a plain grey hoodie, jeans and sneakers stands in a concrete underpass in flat daylight, looking bored. She raises her hand and spins once toward the camera. As she turns, a sweep of golden light wraps around her body and everything changes in one continuous move: the hoodie becomes a tailored black evening gown, her hair falls in polished waves, the concrete wall becomes a glossy marble hall, the flat daylight becomes warm night lighting with bokeh. She completes the spin, the gown flares out and settles, and she looks into the camera with a confident smile. One seamless whip-pan transition at the moment of the spin, cinematic slow motion on the reveal, rich contrast, glossy commercial look. All clothing is plain and unbranded, no logos, no signs or lettering in the background. Do not depict any real brand or trademark. No dialogue, no subtitles. There is no text anywhere in the frame.
Aug 31, 202610
Read the full promptSilver Jacket Dance On The BeatMusic video. A young woman with warm brown skin, high cheekbones, long braided hair pulled into a high ponytail and a wide confident smile dances hard to a beat in an oversized silver jacket and wide black trousers. She is an ordinary anonymous person and looks like nobody famous. Do not depict any real person, actor, singer, celebrity or public figure, and do not reproduce anyone's likeness. Every move lands exactly on the beat and the cuts land on the beat too. Shot one: a dark studio, a hard white spotlight from behind, smoke in the air, she snaps into a pose as the light strobes. Cut on the beat. Shot two: a wet night street, magenta and cyan neon reflected in the puddles, the camera orbits her fast while she spins. Cut on the beat. Shot three: a bright empty concrete hall, flat daylight, low wide angle, she drops into a floor move straight towards the lens. Cut on the beat. Shot four: back in the dark studio, two more dancers step in beside her in matching silver and they hit the final pose together as the light cuts to black. Fast rhythmic editing, punchy contrast, motion blur on the fastest moves, editorial fashion styling. Plain unbranded clothing, no logos, no emblems, no monograms, no brand marks, no readable lettering on the walls, signs or clothes anywhere. Do not depict any real brand or trademark. The only sound in the clip is the rhythm the dancers make themselves in the room: sharp hand claps, foot stomps landing on the concrete floor, the snap and rustle of the fabric as she moves, breaths, and the natural reverb of the empty hall. Dry realistic room recording. There is no music, no song, no melody, no instruments, no backing track, no vocals and no singing. There is no readable text anywhere in the frame.
Aug 31, 202610
Read the full promptFour Presenters, Four LanguagesA bold white sans-serif headline stays burned across the top of the frame for the whole shot. The words are spelled V-O-I-C-E, S-U-B-T-I-T-L-E-S and P-R-E-S-E-N-T-E-R. The headline reads exactly: Change voice, subtitles and presenter. Below it, a scene. The framing, the set and the lighting never change: the same bright modern studio with a soft grey backdrop, the same chest-up framing, the same soft key light. Only the presenter changes. A woman with dark curly hair in a navy blazer stands in frame speaking to the camera with natural lip movement and small hand gestures. A sweep of light passes across the frame and she is instantly replaced, in exactly the same position and pose, by a man with light skin and short blond hair in a grey shirt, still speaking. Another sweep and he is replaced by an older woman with East Asian features and straight black hair in a cream blouse, still speaking. Another sweep and she is replaced by a man with dark skin and a short beard in a black turtleneck, still speaking. All four are ordinary anonymous people and look like nobody famous. Do not depict any real person, actor, celebrity or public figure and do not reproduce anyone's likeness. A clean white rounded subtitle bar sits near the bottom of the frame the whole time. It carries one short caption at a time, in the language of whoever is on screen at that moment, and the caption changes at every swap. While the first presenter is on screen the bar reads exactly Hello, spelled H-E-L-L-O. While the second is on screen it reads exactly Hola, spelled H-O-L-A. While the third is on screen it reads a short greeting written in Japanese characters. While the fourth is on screen it reads exactly Bonjour, spelled B-O-N-J-O-U-R. Each caption is one single word, correctly spelled, centred in the bar, dark text on the white bar, and no other words ever appear in it. Smooth quick dissolves between presenters, shallow depth of field, clean corporate look. Plain unbranded clothing, no logos anywhere. Do not depict any real brand or trademark. Each presenter speaks that same greeting out loud in their own language, one after another: the first speaks English, the second Spanish, the third Japanese, the fourth French. Clear natural speech, accurate lip sync, one calm voice at a time, no music, no background score, no singing. The headline at the top and the one short caption in the bar are the only text in the entire frame.
Aug 31, 202650
Read the full promptSunlit Apartment WalkthroughReal estate film. A smooth gliding camera moves through a bright modern apartment: it starts in a living room with a linen sofa, a soft rug and tall windows, drifts past a wooden coffee table with a vase of dried branches, and continues into an open kitchen with a stone island and warm wood cabinets. Morning light sweeps across the room and slowly warms into golden afternoon light as the camera moves, long shadows stretching over the floor. A woman in a beige knit sweater walks into the kitchen, pours coffee and looks out of the window, relaxed and at home. She has a soft round face, olive skin, dark hair in a loose bun; she is an ordinary anonymous person and looks like nobody famous. Do not depict any real person, actor, celebrity or public figure and do not reproduce anyone's likeness. Calm architectural cinematography, wide lens, steady gimbal motion, natural light, warm neutral palette, shallow depth of field. Everything is plain and unbranded: no logos, no artwork with letters, no readable signs or lettering anywhere. Do not depict any real brand or trademark. No dialogue, no subtitles. There is no text anywhere in the frame.
Aug 31, 202610
Read the full promptPhoto Prints Merge Into A ReelA bold white sans-serif headline stays burned across the top of the frame for the whole shot and reads exactly: Create Shorts, Reels and TikToks. Every word in the headline is an ordinary correctly spelled word and the headline ends with a full stop. Below it, a scene on a dark background lit by blue and magenta neon. It opens on a thick stack of glossy printed photographs floating in the air, slightly fanned out, the top prints lifting and riffling like a shuffled deck of cards. Each photograph plainly shows a different picture: a city street at night, a coffee cup on a table, a person walking, a sunset over the sea. They are ordinary photo prints with thin white borders and nothing is written on any of them. The stack fans wide open, the photographs spin inward and compress together into a single glowing vertical phone-shaped frame in the centre. As they merge, the still pictures start to move and become live video playing edge to edge inside that frame: the same city street, coffee cup, walking person and sunset now in motion, each shot held for a fraction of a second with hard cuts and quick whip transitions. A clean white rounded bar pulses near the bottom over the footage and a bright rounded button appears under it; both are plain blank shapes with completely empty surfaces and no readable letters. The montage lands on the last shot, the button settles into place and the camera comes to rest — the movement finishes on screen and nothing is cut off halfway through. Sleek dark UI aesthetic, neon blue and magenta glow, glossy reflections, energetic social-media pacing. Do not depict any real brand, app icon, trademark or logo. No dialogue. The headline at the top is the only readable text in the entire frame.

Start creating with Seedance 2.0

Seedance 2.0 is available with any Workroom subscription. Pick a plan and start generating.