A good Seedance 2.5 prompt names what each reference controls, gives every shot its own time range, says what must not change, and ends with the sound. The 22 prompts below follow that pattern across eight formats: product ads, dialogue, action, cinematic nature, stylised animation, vertical UGC, food and texture, and transformations. Every one comes with the clip it produced in ChatCut, so you can see what the words actually did before you spend credits on them.
We picked the formats by reading the top-ranking Seedance 2.5 prompt guides. Product ads, landscapes, action and dialogue appear in most of them. Few give the settings and the cost, and almost none say what went wrong. This guide does both.
Key Takeaways
Give each reference one job. “@Image1 fixes the bottle, @Image2 its side, @Image3 its top” kept one product identical across three shots.
Use whole-second ranges for each shot.0-4s, 4-8s, 8-12s produced the cut, the angle and the order we asked for in every multi-shot prompt here.
Write the end state. Liquid that “settles into a still pool”, a tray that folds “in reverse order”, a loop that “returns to the opening wide view”: stating where a shot ends is what made these work.
Separate talking from doing. In the UGC prompt, Etta speaks with empty hands and demonstrates in a separate shot with no speech, which kept both clean.
Draft at 480p. Every new clip here is a 480p draft: 4.84 credits for 10 seconds, against 26.89 at 1080p. Turn the draft you like into the 1080p final.
Room makeover between frames, a product that unfolds and folds back
How these prompts were written and tested
Each prompt follows the order ByteDance recommends in its Seedance 2.5 prompt guide, trimmed to what the shot needs:
References → look → scene → timed shots → what must not change → audio → “No subtitles.”
Every clip on this page was generated with Seedance 2.5 in ChatCut on November 13, 2026, from the prompt printed above it. Seven were made for this guide (prompts 2, 5, 15, 17, 18, 20 and 22) as 480p drafts; the other fifteen come from our Seedance 2.5, Seedance 2.5 Flash and Seedance 2.5 4K model pages, where the large clips were turned into 1080p finals. The reference images for the new prompts were made with GPT Image 2.6 in ChatCut and are shown with each prompt. Nothing was edited after generation except converting the files for the web.
Under each clip, Check says what to look at in your own result and what happened in ours. For the Seedance 2.0 version of this list, see our Seedance 2.0 prompt examples.
Commercial product film. References: @Image1 is the exact coffee bag; preserve its shape, kraft paper texture, illustration and every word of the label, including SLOW MORNING ROASTERS and ETHIOPIA · WASHED. Look: 16:9, warm morning sun with leaf shadows on a stone counter. 0-3s: slow push in on the bag until the label fills the frame, the letters crisp and readable. 3-6s: a hand tips the open bag and dark roasted beans pour onto the counter in slow motion, the label still sharp. Sound: beans clattering, a quiet morning room. No subtitles, no added text.
Demo
Seedance 2.5 in ChatCut, 1080p, 6 s, one product image as @Image1.
Check Read every word of the label in the last frame, not the first. Here both lines stayed readable through the push-in, but the beans pour from a second bag that enters from the top right instead of the one in the reference. If the pour must come from the product itself, say “the same bag tilts forward” and name the hand that tilts it.
2. Press, flip, pour
Use it for a product with a mechanism: a lid, a button, a hinge.
References for prompt 2: @Image1 front, @Image2 side, @Image3 top. Made with GPT Image 2.6 in ChatCut.
References: @Image1 defines the water bottle; @Image2 supplies its side profile; @Image3 supplies its top view. These are three views of one object. Bright morning light on a pale oak kitchen counter. 0-4s: a slow three-quarter push in toward the closed bottle standing next to an empty clear glass. 4-8s: overhead close-up; an adult thumb presses the round cream button once and the lid flips open with a small spring. 8-12s: side view; the hand tilts the bottle and pours a short stream of water into the glass, then sets the bottle down with the lid still open; clear space remains beside the product. Keep the bottle shape, color and lid identical in every shot. No logos, no text, no extra hands. Audio: quiet kitchen air, one soft click, a lid snap, water pouring into glass, a light marimba note at the end. No subtitles.
Demo
Seedance 2.5 in ChatCut, 480p draft, 12 s, 5.80 credits.
Check Follow the cause and effect: thumb on button, then lid open, then pour, in that order and only once. In this take all three happened in their own shot, the bottle kept its colour and loop, and the lid stayed open when it was set down. “These are three views of one object” stops the model treating the side and top views as extra bottles.
3. A travel poster comes alive
Use it to animate an existing design without breaking its layout.
Animated travel poster. References: @Image1 is the exact poster; keep its layout, illustration style, colors, frame and the word LISBOA unchanged and sharp at the bottom. Look: 3:4, flat vintage travel-poster illustration. 0-3s: the yellow tram rolls slowly down the steep street toward the viewer, its bell swinging; pink bougainvillea flowers sway on the wall. 3-6s: sunlight shimmers on the sea in the distance and a few seagulls glide across the sky, while the LISBOA lettering stays fixed and crisp. Sound: a tram bell, gulls, a gentle breeze. No other text.
Demo
Seedance 2.5 in ChatCut, 1080p, 6 s, 3:4, the poster image as @Image1.
Check The lettering and the frame must not move at all. Here the tram rolls forward, gulls arrive and the sea sparkles while LISBOA and the border stay put. Small motions in a few places work better than asking the whole poster to move.
Seedance 2.5 prompts for dialogue and people
Write the exact line in quotes, say who speaks it, and give the listener something to do.
4. One line and a reaction
Use it for a short dramatic beat: an action, a line, a reaction shot.
Intimate drama scene. Look: 16:9, a cluttered watch repair workshop in late afternoon, warm window light, shallow depth of field, lifelike skin and eyes. 0-3s: close-up of an old watchmaker's weathered fingers setting a tiny brass gear into an open pocket watch with tweezers. 3-6s: close-up of his face; he lowers his loupe, his eyes catch the window light, and he says softly, "Listen. It's running again." 6-10s: reverse close-up of his granddaughter, about twenty, holding the watch to her ear; she goes still, then a slow smile spreads and her eyes well up. Sound: a fine ticking, his quiet voice, faint street noise outside. No subtitles, no music.
Demo
Seedance 2.5 in ChatCut, 1080p, 10 s, text only.
Check Turn the sound on and listen for the full line in his voice, then watch that her reverse shot keeps the same room and light. Here the hands, the line and her reaction each landed in their own range, and the reverse shot is framed over his shoulder.
5. Two people, one question
Use it for a two-person exchange with a reverse shot.
References for prompt 5: @Image1 Neri, @Image2 Sol. Made with GPT Image 2.6 in ChatCut.
References: @Image1 fixes Neri and her apricot vest. @Image2 fixes Sol and his turquoise shirt. Bright courtyard beside a community kite-repair bench. One orange hexagonal kite lies flat between them. 0-4s: shared waist-up view; Neri on screen left touches a ribbon and asks, "Does this tail need another knot?" 4-8s: reverse view of Sol on screen right; after a pause he replies, "One is enough. Let the wind do the rest." 8-12s: return to the shared view; Neri chuckles and both inspect the stationary kite. Keep cameras on the courtyard side of the bench, with unchanged ribbon length and clothing. Audio: Neri and Sol take turns in relaxed English, distant rustling leaves, light ribbon flutter, no music. No subtitles.
Demo
Seedance 2.5 in ChatCut, 480p draft, 12 s, 5.80 credits.
Check The two people must not swap sides between shots. “Keep cameras on the courtyard side of the bench” is the plain-English version of the 180-degree rule; here Neri stayed on the left and Sol on the right through the reverse shot, and each spoke only their own line.
6. Start from a still
Use it when you already have the frame you want and need it to move.
Fashion film. Start exactly on the first frame. The two women in long cobalt coats walk their two white greyhounds across the pink sand toward the right; the wind lifts their coats and the dogs' ears; the camera tracks alongside at their pace; their shadows stretch across the sand. Sound: wind, footsteps on sand, dog tags jingling. No subtitles.
Demo
Seedance 2.5 in ChatCut, 1080p, 6 s, generated from one first-frame image.
Check Count the leashes: each woman should hold one dog. A first frame fixes the composition, so the prompt only has to describe movement. Our step-by-step tutorial builds this same scene from plain text to dialogue.
Seedance 2.5 prompts for action and motion
Give fast action a clear direction, one subject and a camera that changes with the beat.
7. Canyon flight
Use it for a chase that moves from first-person to outside views.
Sci-fi chase. Look: 16:9, canyon city at dusk, orange sun against teal haze, anamorphic lens, motion blur, high contrast. 0-4s: first-person view from a hoverbike racing low through a narrow canyon lined with glowing towers, sparks flying as it skims a metal rail. 4-7s: the camera cuts outside to a side tracking shot as the rider in a black helmet banks hard between two hovering cargo ships. 7-10s: wide aerial as the bike bursts out of the canyon over a vast city and the sun flares across the lens. Locks: one rider, same black bike, no text, no logos. Audio: engine whine rising, wind roar, sparks crackle, a deep cinematic hit at the reveal.
Demo
Seedance 2.5 in ChatCut, 1080p, 10 s, text only.
Check Three shots, three distances. The cockpit POV, the outside shot and the aerial all arrived on time; the middle shot came out as a follow from behind rather than a true side view, which is the kind of drift to expect when a shot is only three seconds long.
Vertical skate clip for social. Look: 9:16, sunset over an empty concrete plaza, warm orange backlight, long shadows, fisheye low angle. 0-4s: a skateboarder rolls fast toward the camera. 4-8s: he pops a kickflip right over the lens in slow motion, the board spinning against the sun, and lands clean behind the camera. Locks: one skater, no text, no logos. Audio: wheels rolling, board pop and land, a short lo-fi beat.
Demo
Seedance 2.5 in ChatCut, 480p draft, 8 s, 9:16.
Check Watch the feet and the board separate and meet again. The flip over the lens is clean; the landing happens back in frame instead of behind the camera, so if the exit matters, end the prompt with where the skater finishes.
9. A herd across the lagoon
Use it for a crowd of animals or people moving together.
Nature spectacle. Look: 16:9, sunset over a shallow lagoon, backlit orange spray against deep blue water, telephoto and drone, slow motion. 0-5s: a herd of wild horses gallops through ankle-deep water toward the camera, manes flying, spray lit gold by the low sun. 5-10s: the camera rises into a wide aerial following the herd across the glowing lagoon, their reflections streaking below. Locks: no riders, no people, no text. Audio: thundering hooves in water, splashes, wind, a low cinematic drone.
Demo
Seedance 2.5 in ChatCut, 1080p, 10 s, text only.
Check In crowds, look for horses that merge or legs that disappear in the spray. The telephoto half holds up; in the aerial half the horses are small enough that the model gets away with less detail.
Cinematic Seedance 2.5 prompts
These need the fewest words and the most specific light.
10. An owl lands in the snow
Use it for a wildlife film beat: approach, detail, reveal.
Wildlife film in falling snow. Look: 16:9, frozen pine forest at dawn, cold blue light with a warm sunrise rim, extreme feather detail, shallow depth of field. 0-4s: a great grey owl glides silently through falling snow toward the camera, wings fully spread. 4-7s: it lands on a frosted pine branch, snow bursting off the needles; close-up as frost crystals glitter on its layered grey feathers and its yellow eyes open wide. 7-10s: the camera pulls back and rises to reveal the owl small against a vast snowy forest valley in the first sunlight. Locks: one owl, no people, no text. Audio: soft wingbeats, snow falling from branches, wind, a distant creak of ice.
Demo
Seedance 2.5 in ChatCut, 1080p, 10 s, text only.
Check The landing is the hard part: talons on the branch and snow falling off it at the same moment. This take has the glide, the close-up and the pull-back in order, and the reveal ends with the owl small in the valley as written.
Wildlife close-up. Look: 16:9, savanna at golden hour, warm backlight, long lens, extreme fur detail. A male lion faces into the wind, each strand of his golden mane lifting and catching the light, dust drifting past his amber eyes. Locks: one lion, no text. Audio: wind through grass, a low rumble.
Demo
Seedance 2.5 in ChatCut, 480p draft, 6 s, 2.91 credits.
Check One shot, one action. The mane and the dust move in the same direction and the eyes stay steady. Short single-shot prompts like this are the cheapest way to test a look before writing a longer sequence.
12. Fireworks in her eyes
Use it for an emotional close-up that opens into a big reveal.
Cinematic close-up to spectacle reveal. Look: 16:9, night, warm firework light on skin against deep navy shadows, shallow depth of field, extreme facial detail. 0-4s: extreme close-up of an elderly woman's face, silver hair strands lifting in the breeze, fine wrinkles and skin texture; bursts of gold and crimson fireworks bloom and fade as reflections in her wet eyes, each burst lighting her face for a moment. 4-7s: the camera slowly pulls back past her shoulder; she smiles and one tear slides down her cheek. 7-10s: the camera cranes up over her head to reveal a harbor city at night, giant fireworks exploding over the water and reflecting on the bay. Sound: deep distant booms, crackling sparks, a crowd cheering far away. No subtitles.
Demo
Seedance 2.5 in ChatCut, 1080p, 10 s, text only.
Check Each burst should light her face as it happens, and the crane-up should end on the harbour, not on her. Both happened here; the tear is subtle, so if it matters, give it its own seconds.
Seedance 2.5 prompts for animation styles
Name the medium, then describe what the medium does to motion.
Claymation stop-motion style. Look: 16:9, handmade clay city rooftops at night, visible fingerprints in the clay, warm lamp light, stepped stop-motion movement. A clay red fox sprints across a rooftop and leaps over a gap to the next roof; the camera tracks alongside. Locks: one fox, no text. Audio: playful plucked strings, soft tapping steps.
Demo
Seedance 2.5 in ChatCut, 480p draft, 6 s, 2.91 credits.
Check Clay drifts toward smooth CG. “Visible fingerprints” and “stepped movement” are the words doing the work; here the rooftops kept their handmade look and the fox read as clay, though smoother than real stop-motion. We ran this prompt against Kling 4.0 in our Seedance vs Kling tests.
14. Ink-wash leap
Use it for a painterly style where the medium reacts to the action.
Chinese ink-wash painting style. Look: 16:9, black ink and soft grey washes on rice paper, a few touches of vermilion red, a pale moon, visible brush strokes and paper texture. A red fox drawn in loose brush strokes sprints across tiled rooftops and leaps over a gap to the next roof, ink splashing and spreading as it lands; the camera tracks alongside. Locks: one fox, no text, no calligraphy, no seals. Audio: a plucked guqin note, soft wind, brush swish.
Demo
Seedance 2.5 in ChatCut, 480p draft, 6 s, 2.91 credits.
Check The ink should splash on the landing, not before, and that is where it happened here. Ink-wash prompts tend to attract decorative writing, which is why this one says “no calligraphy, no seals”.
15. A paper-craft loop
Use it for a short clip that has to replay without a visible jump.
Reference for prompt 15: @Image1, the character and the winder. Made with GPT Image 2.6 in ChatCut.
Reference: @Image1 defines the marmot, apron and ribbon winder. Preserve their layered paper construction. 0-3s: wide shot, the marmot cranks a coral ribbon into a tidy coil. 3-7s: detail shot; the coil gently unrolls back across the tabletop, using distinct stepped movement. 7-10s: return to the opening wide view; the marmot restores the crank and ribbon to their initial positions and pauses. Match the opening and ending wide compositions for a replay loop. Keep cut edges visible and the turquoise background fixed. Audio: cardboard clicks, ribbon brushing paper, a three-note toy-piano phrase returning to its opening note. No subtitles.
Demo
Seedance 2.5 in ChatCut, 480p draft, 10 s, 4.84 credits.
Check Put the first and last frames side by side. Ours matched closely in framing, pose and ribbon position, which is what makes it loop. Seedance has no loop setting, so the prompt has to ask for the return; writing the music to “return to its opening note” helps the sound loop too.
Seedance 2.5 prompts for UGC and social video
Vertical, close to the lens, and with the product doing one clear thing.
16. On the beat
Use it for a music-led clip where an audio file sets the rhythm.
Social music clip. References: @Image1 is the woman (curly dark hair, round glasses, red knit sweater); @Image2 is the exact headphones (cream and tan, over-ear); @Audio1 is the music track and sets the beat. Look: 9:16, warm golden wall, late afternoon sun. 0-3s: she lifts the headphones from @Image2 and puts them on, smiling. 3-8s: she closes her eyes and dances to the beat of @Audio1, nodding and swaying, her curls bouncing; the cuts land on the beat. Keep her face and the headphones consistent. No subtitles.
Demo
Seedance 2.5 in ChatCut, 1080p, 8 s, 9:16, two images and one music track as references.
Check Look at the product across cuts. Her face held, but the headphones’ ear cups shift between tan and cream from shot to shot. For a product ad, add “same colours in every shot” and give the product its own close-up.
17. A talking-head demo
Use it for a creator-style ad: a hook, a demonstration, a payoff.
References for prompt 17: @Image1 Etta, @Image2 the phone stand. Made with GPT Image 2.6 in ChatCut.
References: @Image1 fixes Etta and her green shirt. @Image2 fixes the coral folding phone stand and its single hinge. Vertical phone video in a bright home kitchen at midday. 0-4s: Etta talks to the lens with empty hands visible: "Following a recipe with flour on your hands? Here's my fix." 4-8s: cut to her hands on the counter; she unfolds the stand once at its hinge, sets a phone on it, and steps back. No speech during this detail shot. 8-12s: back to Etta, hands away from the stand: "Hands-free, and it doesn't slide." Keep the lowest fifth of the frame uncluttered for captions. No logos, no readable screens. Audio: Etta speaks conversational English; a soft hinge click, a light tap as the phone lands, kitchen ambience, a quiet plucked bass accent after the click. No subtitles.
Check Talking and handling the product in the same shot is where UGC prompts usually break: lips and fingers both have to move right. Splitting them worked here. She speaks with empty hands, the stand unfolds once at its hinge in a silent close-up, and the phone sits on it in the last shot. The bottom of the frame stays clear for captions you add in the editor.
18. First-person carry
Use it for a POV clip where an object must survive being hidden.
An unbroken first-person shot in a sunny sensory garden. Adult hands carry a shallow terracotta tray holding three purple radishes with leaves. Walk beneath a hanging herb arch; foliage briefly hides the tray, then reveals the same three radishes in unchanged positions. Stop at an empty potting shelf, lower the tray once, release it and withdraw both hands. The camera follows at chest height without spinning. Preserve the tray weight during lowering. Audio: the unseen adult carrier says, "Three little treasures for lunch," before setting down the tray; footsteps, brushing leaves and one earthenware contact sound, no score. No subtitles.
Check Count the objects before and after they are hidden. Three radishes went under the leaves and three came out, in the same order; the tray was set down and both hands left the frame. The radishes came out round and deep red, closer to small beets, so name the colour more precisely if it matters.
Seedance 2.5 prompts for food and liquids
Liquids look right when the prompt describes how the flow ends.
Macro food shot. Look: 16:9, dark wood, warm backlight, extreme detail. Thick golden honey drips slowly from a wooden dipper onto a honeycomb, the strands stretching and folding, light glowing through the honey and tiny bubbles inside it. Locks: no text. Audio: soft drip, low warm ambience.
Demo
Seedance 2.5 in ChatCut, 480p draft, 6 s, 2.91 credits.
Check The strand should thin and fold on the comb rather than splash. Here it builds a glossy mound on the cells. This is also our liquid test against Kling 4.0 in the comparison.
20. A spoonful settles
Use it for any pour that has to stop believably.
Reference for prompt 20: @Image1 sets the bowl, the jug and their size, not the framing. Made with GPT Image 2.6 in ChatCut.
Reference: @Image1 defines the blue bowl, jug and their relative scale, not camera framing. 0-3s: oblique close-up, an adult hand lifts the jug and tilts it slowly over the empty bowl. 3-7s: extreme detail, a narrow violet blueberry coulis stream coils into the bowl; briefly slow the apparent motion to reveal glossy folds. 7-10s: wider detail at ordinary speed; the jug turns upright, the stream thins and breaks, one final drop lands, and the coulis settles into a shallow still pool. The liquid transfers only once. Preserve both vessels and avoid added ingredients. Audio: quiet breakfast-room air, a soft pour becoming separate drops, muted glass contact as the jug returns to the table, no music. No subtitles.
Demo
Seedance 2.5 in ChatCut, 480p draft, 10 s, 4.84 credits.
Check Most AI pours never stop. This one coils, thins, breaks and settles into a still pool, with less in the jug at the end than at the start. “Not camera framing” kept the model from copying the overhead angle of the reference.
Seedance 2.5 prompts for transformations
Show both states, then describe the order of change between them.
21. A room between two frames
Use it for a before-and-after with a first and a last frame.
Interior makeover in one shot. The camera stays still with a slow push in. Start exactly on the first frame and end exactly on the last frame. 0-3s: the sunny pink room is calm; the red armchair spins once and turns green. 3-6s: a black and white checkerboard rug unrolls across the wooden floor from the left, and a potted fig tree grows up beside the bookcase. 6-8s: the light settles as the room becomes the last frame. Sound: a soft whoosh, a fabric flutter, a gentle chime. No subtitles.
Demo
Seedance 2.5 in ChatCut, 1080p, 8 s, 4:3, first and last frame.
Check The walls, window and bookcase must not move while the furniture changes. Each change got its own seconds, so they happen one at a time: chair, rug, tree. Frame mode and reference images can’t be combined in one Seedance 2.5 request; pick one.
22. The petal tray
Use it for a product that opens and closes, shown as a playful mechanism.
References for prompt 22: @Image1 closed, @Image2 open. Made with GPT Image 2.6 in ChatCut.
References: @Image1 defines the folded picnic pod. @Image2 defines the reachable open configuration of that same object. Photographic tabletop demonstration in broad daylight. 0-3s: three-quarter view, four hinged panels unfold outward in sequence, each attached by its coral fabric strip. 3-6s: overhead view of the empty open tray, held still. 6-10s: return to the three-quarter view as all four panels fold inward in reverse order, finishing as the original closed pod. This is a playful self-operating mechanism. Keep exactly four panels; no melting, particles or new pieces. Audio: fabric bends, gentle panel taps and light hand-drum pulses slowing to silence. No subtitles.
Demo
Seedance 2.5 in ChatCut, 480p draft, 10 s, 4.84 credits.
Check Count the panels in every shot. There were four throughout, joined by their coral strips, and the tray finished closed as it began. The “overhead” shot came out as a high angle rather than straight down; “top-down, camera pointing straight at the table” is more literal if you need it.
What did these prompts cost?
Clip length
480p draft
720p
1080p
6 s
2.91 credits
6.54 credits
16.18 credits
10 s
4.84 credits
10.87 credits
26.89 credits
12 s
5.80 credits
13.03 credits
32.24 credits
The seven new clips on this page cost 35.79 credits together as 480p drafts, and their reference images 0.15 to 0.31 credits each. At ChatCut’s plan prices of roughly $0.18 to $0.25 per credit, a 10-second 480p draft costs about $0.87 to $1.21. Turning a draft into its 1080p final costs the 1080p price and reuses the prompt, references, length and audio; it must be done within 7 days. The Seedance 2.5 price guide has every duration and resolution.
Three habits kept the bill down: test a new prompt at 480p and 5 to 6 seconds first; change one thing per retry; and only finish the takes you will use.
Common Seedance 2.5 prompt mistakes
Unlabelled references. If you attach three images without saying what each controls, Seedance blends them. Write “@Image1 is…” for each.
Overpacked time ranges. Two actions in a three-second range usually means one of them goes missing. Give each action its own range.
No end state. “Honey drips” can drip forever. “Settles into a still pool” tells the model where to stop.
Talking while doing. Dialogue and precise hand work in the same shot is where faces and fingers break. Put them in separate shots.
Forgetting the sound. Seedance 2.5 always generates audio. If you don’t describe it, you get whatever it guesses, often music you didn’t want.
What is the best prompt structure for Seedance 2.5?
References first, then the look, the scene, each shot with a whole-second time range, what must stay the same, and the audio, ending with "No subtitles." This follows ByteDance's own Seedance 2.5 prompt guide; every prompt on this page uses it.
How do I reference images in a Seedance 2.5 prompt?
Attach the images and refer to them in order as @Image1, @Image2 and so on, saying what each controls, such as "@Image1 fixes the bottle; @Image2 supplies its side view." Seedance 2.5 takes up to 30 images, 10 videos and 10 audio clips as references. A first and last frame are a separate mode and can't be combined with reference images.
How long can a Seedance 2.5 prompt's video be?
4 to 30 seconds per clip. The prompts here run 6 to 12 seconds; longer clips are better split into time ranges of 3 to 5 seconds each.
Do Seedance 2.5 prompts need to describe sound?
Yes. Seedance 2.5 generates audio with every clip, so describe ambience, effects, music and any dialogue with the speaker named. If you don't want music, say "no music".
Can I use these prompts with Seedance 2.0?
The structure works with Seedance 2.0, but 2.0 clips are limited to 15 seconds and 9 image, 3 video and 3 audio references, and it has no 480p draft-to-final route. Our Seedance 2.0 prompt examples are written for it.
