Skip to main content

Seedance 2.5: Prompt Guide

Learn how to direct references, action, dialogue and sound in Seedance 2.5

Prompt Guide

Learn how to direct references, action and sound in Seedance 2.5.

Overview

Seedance 2.5 is ByteDance’s video model. It generates up to 30 seconds at up to 1080p, with sound created alongside the picture. Start from text or use image, video and audio references to guide the result.

How to write the prompt

  • Start with who or what is doing what. Add the setting, look, camera and sound as needed.
  • Give every reference one job: @Image1 for appearance, @Video1 for motion, @Audio1 for voice or music. Say what to borrow and what to leave out.
  • If a video already defines the movement and cuts, tell the model to follow them. Avoid adding conflicting directions.
  • For longer scenes, write stages in playback order, each with one main change and a clear ending. Add timestamps only when timing matters.
  • Keep identities, outfits, props and screen direction consistent across shots. With several characters, map each one to its reference.
  • Describe emotion through visible behavior: a held breath, a dropped gaze, a hand tightening.
  • Name the speaker, language and delivery for dialogue. Give ambience, effects and music their own line.

Prompt structure

Subject + Action + Setting + Visual style + Camera + Audio

Add reference roles first when using them. For a simple shot, a few sentences are enough:

@Image1 defines [appearance and outfit].
@Video1 defines [motion and camera], without its original subject or setting.

[Subject] performs [action] in [setting].
Use [lighting, color and texture] with [framing or camera movement].
Audio: [dialogue, ambience, effects or music].
Keep [identities, wardrobe and props] consistent.

For multiple shots, expand the action into stages: what happens first, what changes next, and where the scene ends.

Build a full edit from one image — Fashion Concept

Use the image for appearance and styling, then describe the scene progression and recurring visual motifs.

Fashion Concept — latest output
Prompt

Create a 30-second high-fashion, surreal, concept-driven fashion film. Use the reference image only for the character’s presence, hairstyle, makeup, and black hemp-rope element.

The lead is a woman with pale luminous skin, a short straight platinum-blonde bob with blunt bangs, cool sharp blue eyes, natural realistic skin texture, and muted red lips. Her presence is calm, dangerous, distant, commanding, fashionable, and mysterious. The black hemp rope must read as avant-garde, elevated, sculptural, conceptual couture—not ordinary restraint or bondage. Different scenes may use different rope looks, but keep the same look within each scene with no abrupt wardrobe changes. Rope styling may include a high-neck structure, exaggerated shoulders, a facial frame, sculpted shoulder-and-neck contours, arm structures, a back piece, dimensional braiding, and asymmetrical silhouettes.

The overall image should feel like Hasselblad medium-format film photography: a luxury fashion-editorial spread, vintage-film atmosphere, soft but clear rendering, detailed realistic skin, subtle film grain, a low-saturation cool palette, layered pure blacks and cool grays, a slight steel-blue undertone, natural refined highlights, shallow depth of field, and elegant bokeh. The pace should be fast and precise. Do not create rhythm by constantly introducing new locations; build variation through changes in camera position, action, expression, flash-lit cuts, and callback shots. Actions include: lifting the eyes, tilting the head, looking back, raising the chin, lightly touching the rope, adjusting the shoulder-and-neck structure, passing a hand in front of the face, stepping forward, making a half turn, lowering the head and looking up again, and standing still while staring into the lens. Expressions include: an indifferent stare, a contemptuous upward glance, slightly parted lips, briefly closing and reopening the eyes, a provocative stare into camera, and an empty yet sacred gaze. The performance must remain elevated, restrained, editorial, and non-casual. Throughout the film, use several rounds of flash-lit cuts and photographic freeze-frame moments, like a sequence of Hasselblad fashion stills coming to life. Flash cuts may move quickly between a frontal view, profile, extreme close-up of the eyes, close-up of the lips, shoulder-and-neck details, hands touching rope, overhead angles, low angles, and mirror-reflection compositions. Do not use flash-style freeze frames in the twin mirror sequence.

Core concept: What restrained her ultimately becomes her.

Scene structure — approximately 3 seconds per scene, 10 scenes total.

Scene 1 | 0–3s — Frontal close-up awakening

An extreme beauty close-up. Black hemp rope frames her shoulders, neck, and the edges of the image. Introduce the first round of flash-lit cuts with several rapid shots: frontal close-up, eye close-up, lip close-up, and profile close-up. Expression progression: indifferent stare → slightly parted lips → eyes briefly close and reopen.

Scene 2 | 3–6s — Twin mirror composition

Against a pure black background, two versions of her stand facing each other, with black rope wrapping their shoulders, necks, and upper bodies. Shots: medium two-shot, left and right profile close-ups, the two figures looking at each other, and one looking into camera while the other looks across. Expressions and movement: slight head tilts, eye movement, short pauses, parted lips, lowering the head and looking back up. No flash-style freeze frames.

Scene 3 | 6–9s — Hands and black rope close-ups

A bright, minimal background. The rope must not look overly smooth; preserve visible fibers, roughness, slight fraying, and authentic hemp texture. Create more interweaving, crossing, knotting, and layered compression between the strands. Shots: open palm, extended fingers, fine rope crossing through the gaps between fingers, overhead hand composition, and close-ups of fingertips and knots.

Scene 4 | 9–12s — Black stage and main installation

A pure-black stage as the primary hero scene. She stands centered in front of a huge, symmetrical, petal-shaped or interlocking circular black-rope installation, like a black floral crown or dark sacred halo. Shots: ultra-symmetrical full-body wide shot, slow push into a medium shot, low-angle view, and a slight overhead view. Actions: head lowered, slowly lifting her head, taking one gentle step forward, then standing still and looking directly into camera.

Scene 5 | 12–15s — Close-up callback

Return to close framing, now from a high angle with her face slightly turned and the black rope positioned closer to her face. Shots: frontal view, profile, overhead angle, eyes, lips, and a hand lightly touching the rope beside her face. Expressions: provocative, empty, contemptuous. Use a denser sequence of flash-lit cuts.

Scene 6 | 15–18s — Main installation callback

Return to the black-stage installation with the same styling and no wardrobe change. Shots: lateral dolly, looking through the installation, top-down overhead shot, and low-angle orbit. Action: she makes a half turn, stops, and looks back at the camera.

Scene 7 | 18–21s — Installation-art space

Cut to a dark gallery-like installation space filled with suspended black hemp-rope structures and numerous cool metallic black flowers. She stands at the center and gently raises one hand. Do not make the ropes rise abruptly; instead, let the surrounding installation change subtly and in an orderly way, such as the flowers slowly opening. Shots: wide shot, low-angle view, overhead view, close-up, and changes in installation, light, and shadow. The entire space should feel as though it is responding to her presence.

Scene 8 | 21–24s — Twin / hands intercut callback

Return to the twin composition and hand close-ups, intercut at a faster pace: the two figures meet each other’s gaze → left figure in profile → hand and rope intertwine → right figure looks into camera → hand glides across rope. A small number of photographic cuts may be used, but do not disrupt the cool, controlled mood of the twin sequence.

Scene 9 | 24–27s — Buildup before the final reveal

Rapidly intercut: frontal close-up stare, full-body main installation, eye close-up, lip close-up, hand touching rope, and the character inside the installation-art space. The edit becomes noticeably faster and the flash-lit cuts more frequent. The audience should feel that she has fully transformed from being “wrapped” into taking control and being crowned.

Scene 10 | 27–30s — Final reveal

A pure black background or deep void, with strong fashion beauty lighting and flash-lit still-image moments. She appears in full; the black hemp rope has become an intensely sculptural conceptual-couture look, like wearable installation art. Hard-cut shots include a frontal cover composition, a medium close-up looking directly into camera, and a hand touching the rope. Avoid large body movements. She lifts her eyes coldly and stares directly into the lens. End on a powerful, frontal, magazine-cover pose.

Try Seedance 2.5

Use video for motion and an image for the look — UniCycle

The video defines movement, camera and edit timing; the image defines the person, outfit, unicycle and color treatment. State those roles explicitly.

UniCycle — latest output
Prompt

[REFERENCE CONTROL] @Video 1 defines ONLY the exact 30-second action and editing blueprint: shot order, cut timing, downhill route, rider trajectory, stunt timing, camera positions, tracking paths, lens changes, foreground wipes, speed changes and audio rhythm. Translate the bicycle performance into physically believable extreme mountain-unicycle riding. Do not inherit the original rider, orange sportswear, helmet, bicycle design, dry brown palette or visual styling. @Image 1 is the canonical identity, fashion, unicycle, environment and color reference. Preserve the same adult man, short dark hair, facial structure, pistachio-green tailored suit, pale-green shirt, cropped trousers and black loafers throughout. No helmet, sportswear or visible protective equipment. Preserve the pictured unicycle’s mint frame, chrome seat post, black saddle, silver rim, single black tire, crank and two pedals. Adapt the tire only with subtle mountain tread; do not otherwise redesign it. Use the image’s saturated emerald landscape, scattered oranges, soft telephoto rendering, restrained editorial mood and exact green-forward color treatment. It is a visual reference, not a first frame. [REFERENCE LOCK] Exactly one man, one unicycle, one wheel and one continuous downhill route. The output must never contain a bicycle, second wheel, handlebar, fork resembling a bicycle, extra rider or duplicate person. Keep the same face, hair, suit, shoes and unicycle from first frame to last. [STYLE AND ENVIRONMENT] A surreal high-fashion extreme-sports film. Transform @Video 1’s exposed alpine route into an intensely green mountain landscape while preserving its exact geography: steep chute, narrow trail, first berm, linked S-turns, three stone steps, cliff straight, natural gap and safe landing shelf. Velvet emerald slopes, pale pistachio grass, deep forest-green shadows and occasional small oranges scattered safely outside the riding line. The oranges remain stationary and never obstruct the rider. Match @Image 1’s grading precisely: layered saturated greens, pistachio tailoring separated subtly from the landscape, warm-neutral skin, lifted blacks, low-to-medium contrast, soft highlight roll-off, restrained halation, fine 35mm grain, gentle motion blur and no digital oversharpening. Elegant deadpan fashion editorial, not conventional sports advertising. [DIRECTOR GRAMMAR] External camera crew only. No POV, helmet camera, chest camera or first-person view. Preserve @Video 1’s exact hard-cut structure and camera choreography. The rider continuously travels from high frame to low frame and generally screen-left to screen-right. Re-establish the downhill direction immediately after every cut. Cut only on wheel impact, body lean or foreground occlusion. 0–2s — Stabilized aerial 70mm extreme wide looking diagonally down the terrifying green ridge. The man enters at the top edge, chooses the steeper chute without stopping and drops immediately. Camera dives to reveal the entire fall line. Bass impact as the single wheel crosses the ridge break. 2–5s — Hard cut to locked 135mm side view across the canyon. Long-lens compression makes the gradient severe. He pedals rapidly through frame while the camera pans down just enough to hold him. A foreground rock wipes the lens. 5–8s — Rock wipe reveals a ground-level 24mm beside the first berm. He leans the unicycle aggressively and carves past within one meter, left to right. Exactly one tire throws one physically sourced fan of gravel backward. Camera shakes once from the pass and holds the empty trail for four frames. 8–12s — Stabilized 24mm chase drone four meters behind and three meters above. It dives at matching speed through two linked S-turns. His pedal cadence matches wheel rotation; hips, knees and arms make controlled balancing corrections while preserving full downhill speed. 12–15s — Cut to a 28mm cable-camera leading him from six meters ahead. He descends three stone steps in one fluid sequence. The single tire contacts each step once; knees and torso compress naturally on each impact. Both feet remain fixed to the two pedals. 15–18s — Hard cut to 85mm lateral long-lens pan across the cliff. He crosses three bands of sun and shadow, leans through one fast berm and releases into the straight. Controlled background streak and one dust ribbon follow the single wheel. 18–21s — Locked ground-level 20mm outside the riding line. He charges directly toward frame, passes safely camera-right at full speed and his pistachio jacket wipes the image. Wind and tire sound peak without camera collision. 21–25s — The pistachio wipe opens to a 35mm aerial side-wide traveling parallel. He accelerates briefly to 115%, compresses downward on the pedals, preload-hops from the natural lip and clears one real gap from left to right. At takeoff, music drops to sub-bass and wind. While airborne, use one 45% slow-motion beat as the drone arcs 30 degrees clockwise. Return to real time before the tire reaches the landing. 25–28s — Cut on single-tire impact to a low 50mm side follow. The tire compresses visibly, his knees and torso absorb the landing and his arms correct balance. One dust burst travels backward. He stays seated over the wheel and accelerates down the final straight. 28–30s — Locked 100mm telephoto at the bottom. He races through frame and exits screen-right without slowing. Music and tire sound cut sharply on exit. Hold the empty emerald trail and drifting dust for the final half-second.

Music and tire sound cut sharply on exit. Hold the empty emerald trail and drifting dust for the final half-second. [PHYSICS] All riding must be mechanically coherent for an expert mountain unicyclist. Exactly one wheel at all times. Both feet remain attached to the two pedals. Pedal cadence and crank rotation match wheel speed. The saddle and mint frame remain centered beneath him. Jumps use visible preload from hips, knees and pedals. Landings occur on the single tire with believable compression, balance correction, gravity and momentum. [AUDIO] Original 150 BPM instrumental: tight breakbeat drums, distorted sub-bass, short metallic percussion and granular synth, without vocals or recognizable melody. Lock beats to berm pressure, stair impacts, takeoff and landing. Layer mountain wind, one-wheel tire roar, gravel spray, pedal and crank clicks, close-pass Doppler, fabric movement and landing thump. No bicycle chain or freehub sounds. No dialogue or voiceover. [FINAL LOCKS] Same man, same pistachio suit and same mint unicycle throughout. EXACTLY ONE WHEEL. No bicycle, second wheel, handlebar, helmet, sportswear, wardrobe change, duplicated pedals, detached feet, malformed limbs, equipment morphing, extra people, moving oranges, logos, captions or on-screen text. Only the specified airborne slow-motion beat; everything else runs at full believable speed. End after the rider exits and the dust settles.

Try Seedance 2.5

Let the soundtrack drive the scene — Anxiety MV

Make the audio reference the timing authority, then tie choreography, cuts and lip-sync to the song.

Anxiety MV — latest output
Prompt

Reference mapping - Images 1-7 define seven separate adult dancers: exact identity, face, hair, body proportions, personality, OOTD and footwear. Image 1 is the lead performer. Never duplicate, merge, replace or reorder the cast. - prompt-imgImage 8 defines the empty daylight warehouse: dark concrete floor, raw industrial walls and one diagonal warm sunbeam. prompt-imgImage 9 defines the wet cobalt city street: centered yellow lane lines, deep blue night ambience and reflective pavement. prompt-imgImage 10 defines the symmetrical tomato-red performance chamber: red doors, metal wall rails, red-and-black checkerboard floor and hard rectangular ceiling panels. Preserve each location's exact architecture, floor direction and lighting; never blend the three worlds. -

Audio 1 is the supplied 30-second master “Anxiety Behind Me.” It is the sole timing, music, lyric and performance authority. Global brief Create one exact 30-second, 16:9 hip-hop performance film in which the same signature eight-count grows from one dancer to seven and transfers cleanly through three worlds. The music drives every chest hit, head snap, foot landing, formation change, camera cut and lighting cue. The verse feels physically trapped and restless; the hook releases into a bigger pursuit formation; the outro resolves through one visible breath. Use energetic shoulder-mounted and handheld coverage with human operator drift, quick reframing and close physical proximity, while retaining one readable full-body master in every technical dance phrase. Alternate intimate face, eyes, hands, chest and footwear close-ups with low full-body proof. Every close-up is motivated by a lyric, impact or breath; never hide a landing or cut mid-word. Visual and art style Contemporary Y2K street-performance editorial: charcoal, bone, faded red and cobalt wardrobe; diagonal warehouse daylight, wet cobalt street reflections and tomato-red checkerboard performance light; natural matte skin, bright readable eyes, fine 35mm grain, mild broadcast texture and crisp body silhouettes. Alternate low 18–28mm handheld full-body movement with intimate 50–100mm identity, texture and vocal close-ups. Preserve controlled natural shoulder drift and breathing proximity without chaotic shake. No glossy idol lighting, oily skin, crushed faces, decorative slow motion, floating bodies or excessive cutting. Signature eight-count Count 1: right palm strikes the sternum on the kick. Count 2: left forearm cuts across the face. Count 3: head snaps screen-right. Count 4: left foot drags outward. Count 5: both elbows rebound behind the ribs. Count 6: one compact body roll drops the center of gravity. Count 7: right knee catches upward. Count 8: both feet land wide and hold. Preserve the order and recognizable silhouette every time the phrase returns; later versions may change formation and travel but not the movement identity. Shot and audio timeline 0–5.53s — Verse A, warehouse solo. Lyrics: “Heart racing, mind pacing, can't sit in one place / Ten tabs open in my head, every thought's a race.” Start low 24mm handheld full-body on the male lead alone inside the warm sunbeam, both shoes visible. The operator creeps forward with restrained shoulder drift as he performs the exact eight-count. On “Heart,” cut after impact to an 85mm handheld insert of his palm compressing the jacket at his sternum; on “mind,” a 65mm three-quarter face close-up catches the forearm passing his eyes and the head snap. Return to full-body before the left-foot drag and keep count 8 visible after “race.” He does not lip-sync; his mouth remains relaxed and closed. Cut only after both feet plant. 5.53–10.75s — Verse B, warehouse crew reveal. Lyrics: “Clock tickin' loud, but I'm frozen in time / Counting all the what-ifs, they be stackin' in line.” Scale-contrast hard cut to a high 28mm handheld wide: six dancers already form a shallow arrow behind the male lead. The camera descends slightly and drifts laterally with the formation centroid while the lead repeats the phrase and dancers 2–7 inherit it one count apart. After the complete phrase “frozen in time,” one brief 100mm handheld close-up catches his held eyes and jaw while the final dancer remains legible behind him completing count 7. Return wide before “stackin' in line,” where all seven land one clean diagonal. No lip movement; every foot stays readable. 10.75–15.96s — Verse C, wet-street pressure. Lyrics: “Phone buzzin', ceiling starin' back at me / Try to breathe it out but it won't let me be.” Pose-match the diagonal to the wet cobalt street with the male lead centered over the yellow line. A low 24mm shoulder-mounted lateral track follows three grounded steps while the crew shadows him one pace behind. On “buzzin',” cut after the shoulder accents to an 85mm jacket, neck and breath close-up; on “breathe,” move into a close 50mm handheld face-and-chest frame as he opens his ribcage and the six bodies close around him in soft depth. His mouth stays closed. Pull back to full-body before the final compression and cut in the real gap at 15.96s. 15.96–21.19s — Hook A, street pursuit. Lyrics: “Anxiety, ridin' right behind me / Try to lose it but it always finds me.” Start on an intimate 85mm handheld chest-up frame of the male lead walking toward camera; the operator backs up at his breathing pace with controlled natural sway. He alone lip-syncs exactly, with every phoneme, eye movement and breath sharp and unobstructed. At the completed line break, cut to one 100mm eye-and-mouth close-up with wet cobalt bokeh while the six backing dancers advance softly behind him with closed mouths. After “finds me” finishes, rapidly but smoothly retreat to a low 24mm full-body wide; all seven perform the eight-count in unison and travel two grounded steps toward camera. End count 8 centered. 21.19–26.39s — Hook B, red-chamber peak. Lyrics: “Whisper in my ear when the room get quiet / Turn a little calm into a full-on riot.” Pose-match count 8 into the tomato-red chamber. Begin 18mm low handheld wide with the checkerboard and ceiling panels converging on the male lead. He continues exact lip-sync while the crew performs mirrored rows with closed mouths. After the complete word “whisper,” cut to an 85mm handheld profile as one dancer leans near his ear without contact; make a short 30-degree orbit around his face, catching eye tension and jacket texture, then return wide before “Turn.” On “riot,” the group bursts by real steps from tight column to a seven-person fan as the ceiling panels pulse once center-to-edge. Push toward his face only after all seven feet plant. 26.39–28.76s — Outro breath. Lyric: “Breathe in, breathe out — tellin'.” Use a locked-but-breathing 65mm handheld frontal close-up on the male lead. He alone lip-syncs the audible words; shoulders rise on “Breathe in” and settle on “breathe out.” Focus stays on his eyes while the six dancers lower their arms behind him with closed mouths. At the last syllable, his mouth closes and the operator's movement settles with his gaze. 28.76–30s — Silent landing. Cut to a locked 28mm full-group wide in the same red chamber. All seven hold the final formation without lip movement. One residual ceiling-light pulse fades, fabric and breathing settle naturally, and the image remains still through 30.00s. Audio and vocal rules Use Audio 1 exactly from 0.00–30.00s. Preserve its original tempo, mix, lyrics, breaths, brief 15.96s section gap and silent final tail. Do not replace, remix, time-stretch, loop or generate new music. No voice-over, extra dialogue, translated lyrics, ad-libs, crowd vocals or added sound effects. The lead alone lip-syncs during 15.96–28.76s; all other mouths remain naturally closed. Never cut a face close-up in the middle of a word. Continuity and restrictions Preserve all seven identities, OOTDs, footwear and formation ownership. Maintain lead at visual center through every world transfer. Feet contact the ground; knees load before direction changes; hair and fabric settle after each landing. No extra dancers, identity swaps, duplicated bodies, impossible camera teleportation, blended locations, subtitles, captions, logos, microphones or handheld props.

Try Seedance 2.5

Keep a large cast readable — Backstage

Map the cast to references, limit who appears in each shot, and carry one character’s emotional arc through the scene.

Backstage — latest output
Prompt

SCENE BACKGROUND The last thirty seconds before ten performers walk onstage. A backstage corridor and wing, never the stage. One performer moves from private panic to composure and pulls the group together. Every shot is handheld: violent in the opening, a slow push in the private beat, tracking her feet in the leader section.

ACTIVE REFERENCES All ten references are character sheets, one performer each - exact face, hair and costume locks, never merged. THE LEADER, who carries the arc and holds the centre, is the FIRST character sheet. There are no location or prop references: the corridor, wing and doors are built from this description alone.

IDENTITY LOAD RULE No more than THREE performers visible in any frame except the assembly at 18.0-23.0s, the only shot containing all ten. Reaction shots are singles or pairs, close, each face large. The film ends on one face, not ten.

SHOT AND ACTION TIMELINE [0.0-6.0s - CHAOS - three quick shots, violent handheld] Hard cuts every two seconds. One performer tears through an open case hunting a missing accessory, throwing things aside. One holds a broken skirt zip open at the waist, gesturing for help. Third shot wider: bodies crossing the corridor both ways, a rail shoved past camera, nobody in control.

[6.0s - a single-frame white flash cuts the chaos dead with a ringing sound, like a 7000Hz tinnitus ringing tone]

[6.0-12.0s - THE PRIVATE MOMENT - slow push-in, then two ECUs] The leader alone at the end of the corridor, back to the group, forehead almost touching the wall. Group noise drops to muffled and out of focus behind her. 18-degree ECU on her hands clutched together, knuckles pale. 18-degree ECU on her lips as she breathes out once, long and deliberate. The camera rises from her lips to her eyes and holds - the panic gone, ringing tone fades out, something decided in its place.

[12.0s - HARD MATCH CUT on the rise of her head]

[12.0-18.0s - THE LEADER - handheld tracking out, following her feet] She is already walking, completely composed, chin level, a different person from six seconds ago. The camera tracks backwards ahead of her feet then rises to her face. She claps twice, sharp. THE LEADER (bright, in command, in English): "All right, girls. Time to put on a show." Cut to three quick reaction shots at different angles, one or two performers each: heads coming up from their chaos, a mirror abandoned mid-touch-up, a grin breaking.

[18.0-23.0s - THE ASSEMBLY - 65-degree wide, handheld, rising] The only shot containing all ten. They arrive from different directions and fall in together, walking toward the doors slightly out of step in a staggered manner - not choreographed, real people moving as a group. Exactly TEN distinct performers, each face separately readable, each costume as referenced. Count them: ten. Hard cut.

[23.0-26.0s - HANDS - 18-degree macro, tight, no faces] Hands stack into the centre one at a time, ten of them, the leader's last, pressing down. Sleeves, rings and mic packs read in detail. Breathing synchronises audibly.

[26.0-28.0s - THE DOORS - 65-degree wide, low, from behind the group] The doors begin to part. Stage light enters as a widening vertical blade and blows out the exposure before it settles. Crowd sound rises with the gap. The ten stand in formation, backs to camera, the leader at centre. Hard cut.

[28.0-30.0s - HER FACE - 85mm, handheld, tight on THE LEADER at centre] Reverse to the front. Stage light crosses her face from directly ahead, catching her eyes. She does not move and does not step through. Her expression holds two things at once - nervous and completely certain. Micro-movement only: one blink, breath. Hold to the end. The film ends on her, not the stage.

PHYSICS Costumes carry inertia and settle a beat after each body stops. The clap is sharp and real. The doors part with mechanical weight and the air pressure changes audibly.

LIGHTING Lit only by practicals - 3200K work lights and mirror bulbs, uneven and unglamorous. The one dramatic event is at 26.0s when 5600K stage light enters through the opening doors, and at 28.0s becomes a hard frontal key on her face alone.

AUDIO No music, no score. Opening: urgent overlapping voices, a case emptied, hangers on a rail, tape tearing. At 6.0s everything drops to muffled room tone behind one long controlled exhale. From 12.0s sound returns with her clap and voice, then mic packs and muffled crowd through the doors. At 26.0s the crowd rises to full volume. The final two seconds are crowd noise and her breath only. One spoken line, hers. No singing.

POSITIVE LOCK All ten identities stay 100% locked to their sheets, costumes exactly as referenced. Only these ten appear. The camera never passes the doors. The stage itself is never seen. Nobody sings and nobody performs.

STYLE Observational documentary: handheld, practical light, honest skin texture, fine grain, restrained grade. No on-screen text, no logos, no watermark.

Try Seedance 2.5

Direct a quiet emotional turn — TV Drama

Set the mood, write the scene beats, then describe the small reactions and camera moves that carry the emotion.

TV Drama — latest output
Prompt

[Style and Mood]

Medium-paced Japanese TV-drama narrative pacing, cinematic winter-romance aesthetic.

Plot advances at television-drama density while holding a restrained blue-cyan snow palette, warm cream interior lighting, natural warm skin tones, restrained Fuji-film grain, and a light softening across the image.

[Narrative]

She steps into a convenience store in Tokyo on a winter night, snow still on her shoulders. Inside, the warm cream light breaks the cold blue outside. She moves down an aisle, unhurried, reaching for something ordinary — and stops. Across the store, half turned toward the drink fridge, is her first love. Neither moves for a moment. A flicker of recognition crosses her face before she can control it — surprise, then something older and quieter underneath.

[Scene Beats]

  1. Exterior — she approaches the convenience store through falling snow, the sign’s light glowing ahead.
  2. She enters; the door chime sounds; warm interior light replaces the cold blue night on her face.
  3. She browses an aisle, ordinary and unbothered, reaching for an item.
  4. She looks up and sees him — a beat of stillness, held just long enough to register as history, not coincidence.
  5. A wordless exchange: a small nod, a half-step forward that goes nowhere, the distance between them staying exactly the same.
  6. She looks away first, composes herself, and continues toward the counter — the moment already receding into something she will carry out into the snow.

[Camera and Performance]

Handheld-adjacent stability, natural TV-drama coverage — a slow push on her face at the moment of recognition, a held two-shot across the aisle, no rapid cuts. Her performance stays restrained: controlled breath, minimal visible reaction, the emotion carried in stillness rather than expression.

Try Seedance 2.5

Start creating

Now make your own

Put this guide to work in Pika.