Table des matières
Seedance 2.5 réalise quelque chose qui semble encore un peu irréel : jusqu’à 30 secondes de vidéo avec son, en un seul plan continu. Sans assemblage ni raccord visible.
Mais un plan long ne pardonne rien. Chaque seconde doit raconter quelque chose, sinon le modèle commence à improviser. Chaque prompt Seedance 2.5 ci-dessous est conçu pour garder le cap pendant les 30 secondes. Ils sont classés par usage : trouvez celui qui se rapproche de votre idée, collez-le et modifiez les détails.
Tous ces prompts fonctionnent dans le générateur vidéo IA Seedance 2.5 sur AI Studio, où vous pouvez utiliser Seedance 2.5 directement dans votre navigateur.
Les nouveautés de Seedance 2.5
ByteDance a lancé Seedance 2.5 le 31 juillet 2026. Certaines de ses fonctions changent la façon de rédiger vos prompts :
- Jusqu’à 30 secondes par génération, avec audio. La version précédente était limitée à 15 secondes par génération.
- Prompts avec repères temporels. ByteDance indique que vous pouvez contrôler l’histoire, la caméra, le mouvement et le rythme sur des plages précises, comme « 0–5s: … ».
- De nombreuses références. Selon ByteDance, jusqu’à 30 images, 10 clips vidéo et 10 clips audio en une génération.
- De meilleures transitions entre plans, et moins de sous-titres et de musique de fond indésirables, selon ByteDance.
- Des points faibles connus. ByteDance reconnaît des améliorations possibles pour la physique des mouvements complexes et la stabilité des scènes où plusieurs sujets interagissent.
Sept habitudes pour de meilleurs prompts Seedance 2.5
- Écrivez par segments temporels. Au-delà d’environ 10 secondes, découpez le prompt en séquences : « 0–8s », « 8–18s », « 18–30s ». Les exemples de ByteDance suivent ce format.
- Nommez vos références. Si vous importez plusieurs images, précisez ce que représente chacune : « la femme de l’image de référence 1, le sac à dos rouge de l’image de référence 2 ». Les exemples de ByteDance utilisent des balises comme @Image 1. Sur AI Studio, de simples descriptions comme celles-ci fonctionnent bien.
- Demandez un plan continu lorsque vous en voulez un. Des expressions comme « single continuous take, no cuts » aident à éviter les coupures.
- Décrivez le mouvement de caméra simplement. « Slow push-in », « orbit right », « handheld follow », « crane up ».
- Ajoutez une ligne pour le son. L’audio est généré avec l’image : précisez ce que vous voulez entendre ou non, par exemple « no music », « no on-screen text ».
- Limitez le nombre de personnages. Le modèle ayant des difficultés avec les sujets en interaction, mieux vaut rester entre un et trois personnages.
- Testez à moindre coût, finalisez en haute qualité. Sur AI Studio, générez les essais en 480p et rendez la version retenue en 1080p.
Le modèle de prompt
[Length], [aspect ratio], single continuous take.
[Subject and look] [action] in [setting, time, weather].
Camera: [shot size and movement].
Style: [light, palette, look].
Sound: [ambience, dialogue, music, or "no music"].
0–Xs: [beat one]
X–Ys: [beat two]
Y–30s: [beat three]
Produit et e-commerce
1. Parfum dans le vent du désert
20 seconds, 9:16, single continuous take. An amber glass perfume bottle stands on a sand dune at golden hour as wind lifts fine sand around it.
Camera: low angle, slow 90-degree orbit, ending close on the bottle.
Style: warm gold and copper tones, glowing backlight, luxury commercial look.
Sound: soft wind, a low, breathy synth tone, no voiceover.
0–8s: Sand streams past the bottle in slow motion.
8–15s: The orbit catches the sun through the glass, throwing amber light across the sand.
15–20s: The wind drops, and the sand settles into perfect ripples.
2. Casque sans fil dans les transports
15 seconds, 16:9, single continuous take. A young man in a charcoal coat wears matte white over-ear headphones on a crowded subway platform.
Camera: medium shot, slow push-in toward his face.
Style: cool blue platform lighting with warm highlights on his face, crisp tech-ad look.
Sound: busy platform noise that fades almost to silence as he adjusts the headphones, then a soft piano melody.
0–6s: A train roars in behind him; the crowd rushes past.
6–15s: He touches the ear cup, closes his eyes, and the noise drops away to calm music.
3. Sac à dos à l’épreuve des sentiers
25 seconds, 16:9. A hiker carries a forest-green backpack up a rocky mountain trail in the early morning.
Camera: handheld follow from behind, then a side tracking shot, ending in a close-up of the pack's buckles.
Style: crisp alpine light, cool blues and greens, outdoor documentary look.
Sound: boots on rock, wind, a small stream, no music.
0–10s: Follow shot up a steep section of trail.
10–18s: Side tracking shot as she steps across a stream on stones.
18–25s: She stops at a viewpoint, clicks the chest strap, and looks out across the valley.
4. ASMR de clavier mécanique
12 seconds, 16:9, single continuous take. Hands type on a mechanical keyboard with cream-colored keycaps on a walnut desk at night.
Camera: macro close-up, slow slide along the keys.
Style: warm desk lamp, soft bokeh from a monitor glow, cozy.
Sound: crisp, deep key clacks, quiet rain on the window, no music.
0–12s: Fingers type steadily as the camera glides from left to right across the keyboard.
5. Crème solaire à la plage
15 seconds, 9:16. A woman in a straw hat applies sunscreen from a pale blue tube on a sunny beach.
Camera: close-up on hands and tube, then a pull-back to a medium shot.
Style: bright summer light, turquoise and sand tones, fresh lifestyle look.
Sound: waves, gulls, distant laughter, a light ukulele tune.
0–6s: Close-up as she squeezes lotion onto her palm.
6–15s: The camera pulls back as she smiles, puts on sunglasses, and walks toward the water.
Cuisine et boissons
6. Sirop sur des pancakes
10 seconds, 1:1, single continuous take. Maple syrup pours onto a tall stack of fluffy pancakes topped with butter and blueberries.
Camera: side angle, slow motion, slight push-in.
Style: warm morning light, golden browns, appetizing food-commercial look.
Sound: a soft drizzle, gentle café ambience.
0–10s: Syrup cascades over the edge and the butter slowly melts down the side.
7. Pizza au feu de bois
20 seconds, 16:9. A pizzaiolo slides a pizza into a glowing wood-fired oven and pulls it out bubbling.
Camera: medium shot at oven height, then a close-up as the pizza comes out.
Style: firelight orange, dark brick, rustic and warm.
Sound: crackling fire, the scrape of the wooden peel, restaurant chatter.
0–7s: He stretches the dough, adds basil, and slides the pizza in.
7–14s: Close on the oven mouth as the crust puffs and blisters.
14–20s: He pulls it out, steam rising, cheese bubbling.
8. Rituel du matcha
15 seconds, 16:9, single continuous take. Hands whisk bright green matcha in a handmade ceramic bowl on a low wooden table.
Camera: top-down view, then a slow tilt to a side view.
Style: soft, diffused daylight, muted earth tones, calm and minimal.
Sound: the bamboo whisk, a kettle, birds outside, no music.
0–8s: Top-down as the whisk turns the matcha into a fine froth.
8–15s: The camera tilts as the hands lift the bowl toward the lens.
Mode et beauté
9. Trench-coat dans le vent
15 seconds, 9:16, single continuous take. A model in a camel trench coat walks across a city bridge on a windy, overcast day.
Camera: slow-motion tracking shot from the front, walking backward with her.
Style: muted grey city, the coat as the only warm color, editorial fashion look.
Sound: wind, footsteps, a distant tram bell, minimal electronic beat.
0–15s: She walks toward the camera as the coat billows dramatically behind her.
10. Échantillon de rouge à lèvres
8 seconds, 1:1, single continuous take. A deep red lipstick glides across a white surface in a single swatch.
Camera: macro, top-down, locked off.
Style: clean white studio light, rich saturated red, beauty commercial look.
Sound: a soft, creamy glide, a single chime at the end.
0–8s: The lipstick draws a smooth swatch from left to right, catching a highlight.
11. Veste en jean et style urbain
12 seconds, 9:16. A skater in an oversized vintage denim jacket leans against a graffiti wall, then pushes off on a board.
Camera: medium shot, then a quick whip pan to follow.
Style: sunny afternoon, punchy colors, 1990s street-style film look.
Sound: street noise, board wheels, an upbeat boom-bap beat.
0–6s: He adjusts his collar and looks at the camera.
6–12s: He drops the board and rolls out of frame, and the camera whips to follow.
Voyages et lieux
12. Fenêtre de train et vallée montagneuse
20 seconds, 16:9, single continuous take. View from a moving train window as it winds through a green mountain valley with waterfalls.
Camera: steady, as if resting on the windowsill; reflections faintly visible in the glass.
Style: soft morning light, lush greens, peaceful travel look.
Sound: train rhythm on the rails, a quiet carriage, no music.
0–20s: Meadows, wooden chalets, and waterfalls slide past as the train enters a sunlit stretch.
13. Tramway jaune dans une rue pentue
15 seconds, 16:9. A vintage yellow tram climbs a steep, narrow street of tiled buildings in a hilly old city.
Camera: low angle at the bottom of the hill, slow tilt up as it passes.
Style: bright afternoon sun, blue and white tiles, warm nostalgic tones.
Sound: the tram's bell, metal wheels on rails, street conversation.
0–8s: The tram approaches up the hill toward the camera.
8–15s: It passes close by; the camera tilts to follow it toward the top of the hill.
14. Cabane sous les aurores boréales
25 seconds, 16:9, single continuous take. A small wooden cabin with a glowing window sits in a snowy field under the northern lights.
Camera: slow push toward the cabin from a distance.
Style: deep night blue, swirling green aurora, warm amber window, magical realism.
Sound: soft wind, snow crunching, a faint crackle from a fireplace.
0–12s: The aurora ripples across the sky as the camera moves forward.
12–20s: Smoke curls up from the chimney.
20–25s: The cabin door opens and warm light spills across the snow.
Récit et cinéma
15. Violoniste dans le métro
30 seconds, 16:9, single continuous take. A young violinist plays alone in a tiled subway corridor as people hurry past.
Camera: slow, continuous dolly toward her from the far end of the corridor.
Style: cool fluorescent light, warm light on her face, cinematic realism.
Sound: footsteps and train rumble, then a slow, beautiful violin melody that grows louder.
0–10s: Commuters rush past without looking; the music is faint.
10–20s: The camera moves closer; one child stops and stares.
20–30s: A small crowd gathers in silence as she plays the final note.
16. Robot perdu sous la pluie
20 seconds, 16:9. A small, rusty delivery robot with one flickering headlight rolls down a rainy neon street, looking for an address.
Camera: low tracking shot beside the robot, ending in a close-up.
Style: neon reflections in puddles, pink and blue palette, gentle science fiction.
Sound: rain, distant traffic, soft electronic beeps from the robot.
0–8s: It stops at a doorway and scans the number, beeping uncertainly.
8–15s: It rolls on, its headlight flickering.
15–20s: A door opens and a little girl holds an umbrella over it.
17. La lettre
15 seconds, 16:9, single continuous take. An older man sits at a kitchen table reading a handwritten letter, his tea going cold.
Camera: slow push-in from medium shot to close-up.
Style: soft grey morning light, quiet, intimate drama.
Sound: a clock ticking, birds outside, paper rustling, no music.
0–8s: He reads, his expression unreadable.
8–15s: His eyes fill with tears, then he smiles and sets the letter down gently.
18. Cambriolage au musée
20 seconds, 21:9. A thief in black slips between red security lasers in a dark museum gallery toward a glowing glass case.
Camera: overhead shot, then a low side angle.
Style: high contrast, red laser light, sleek heist-movie look.
Sound: tense low strings, a faint hum from the lasers, careful breathing.
0–8s: Overhead view as she bends backward under a laser beam.
8–15s: Low angle as she steps over the last beam, inches from the case.
15–20s: She reaches the case and smiles, and then an alarm light starts to spin.
Dialogues et personnages
19. Extrait de podcast
15 seconds, 9:16. Two podcast hosts sit at a table with microphones in a cozy studio with acoustic panels and plants.
Camera: medium two-shot, slight push-in.
Style: warm practical lights, natural colors, social clip look.
Sound: clear studio voices, light laughter.
Host 1: "So you're telling me you've never seen the ocean?"
Host 2: "I've seen pictures. That counts, right?"
0–15s: Host 1 asks the question in disbelief, and Host 2 answers with a shrug as both burst out laughing.
20. Démonstration d’un professeur de sciences
12 seconds, 16:9. A science teacher with rolled-up sleeves stands behind a lab bench and holds up a beaker of blue liquid.
Camera: medium shot, then a close-up on the beaker.
Style: bright classroom, clean and friendly, educational video look.
Sound: classroom ambience, a fizz, a few kids saying "whoa."
Teacher: "Watch closely. Don't blink."
0–6s: He says the line with a grin.
6–12s: He drops in a tablet and the liquid foams up and turns bright purple.
21. Moment gênant dans l’ascenseur
15 seconds, 16:9, single continuous take. Two coworkers stand in an elevator holding identical coffee cups, avoiding eye contact.
Camera: static wide shot from the corner, slightly high.
Style: flat office lighting, muted colors, deadpan comedy.
Sound: elevator hum, a ding, faint elevator music.
Coworker 1: "...Nice cup."
Coworker 2: "Thanks. You too."
0–15s: Long silence, then the exchange; the doors open and both step out at the same time and bump shoulders.
Action et sport
22. Descente en VTT
15 seconds, 16:9. A mountain biker in a bright orange jersey rides down a steep forest trail over roots and rocks.
Camera: chest-mounted point-of-view shot, then a trailside angle for a jump.
Style: dappled forest light, rich greens, high-energy action look.
Sound: tires on dirt, rattling bike, heavy breathing, wind.
0–10s: Fast POV descent weaving between trees.
10–15s: Trailside slow-motion shot as the rider clears a small jump.
23. Boxe au petit matin
20 seconds, 16:9. A boxer hits a heavy bag in an old gym before sunrise.
Camera: slow circling shot around the boxer and bag.
Style: a single shaft of dusty light from a high window, gritty, warm tones.
Sound: punches thudding into leather, chains rattling, steady breathing.
0–10s: Steady combinations, sweat flying in the light.
10–20s: She stops, catches her breath, and taps her gloves together, focused.
24. Chien et frisbee
10 seconds, 16:9, single continuous take. A border collie sprints across a green park and leaps to catch a red frisbee.
Camera: low tracking shot alongside the dog, in slow motion for the catch.
Style: sunny afternoon, vivid colors, joyful.
Sound: running paws, a happy bark, people cheering.
0–10s: The dog chases the frisbee, launches into the air, and snatches it mid-flight.
Musique et danse
25. Danseur de rue au crépuscule
20 seconds, 9:16, single continuous take. A dancer in loose white clothes performs a fluid contemporary routine in an empty parking lot at dusk.
Camera: slow orbit around the dancer at waist height.
Style: purple-orange sky, streetlights flickering on, cinematic music-video look.
Sound: a slow, emotional electronic track with a heavy beat.
0–20s: The dancer moves in sync with the beat, spinning, sliding, and freezing on the final drop.
26. Groupe dans un garage
15 seconds, 16:9. A three-piece band (drums, bass, guitar) rehearses in a cluttered garage with fairy lights.
Camera: handheld, moving between members.
Style: warm tungsten light, a little haze, indie music video look.
Sound: an energetic garage-rock riff, cymbals, a shouted "one more time!"
0–8s: The camera weaves from the drummer to the bassist.
8–15s: The guitarist hits the final chord and jumps as the lights flicker.
Contenu pour réseaux sociaux et créateurs
27. Préparez-vous avec moi
15 seconds, 9:16. A young woman gets ready in front of a mirror with round bulbs, doing her makeup and talking to the camera.
Camera: phone propped on a vanity, front-facing, natural.
Style: soft vanity light, natural skin, authentic creator look.
Sound: casual voice, room tone.
Creator: "Okay, five-minute makeup, because I'm late. Again."
0–15s: She speaks the line, swipes on mascara and blush quickly, then grabs her bag and leaves the frame.
28. Découverte d’une chambre transformée
20 seconds, 16:9, single continuous take. A plain, empty bedroom transforms into a cozy, plant-filled bohemian room.
Camera: static wide shot from the doorway.
Style: bright daylight, natural colors; the change happens smoothly.
Sound: an upbeat acoustic track, small "pop" sounds as each object appears.
0–6s: Bare white room.
6–16s: A rug, bed, plants, and string lights appear one by one.
16–20s: Golden hour light fills the finished room.
Utiliser des références
29. Votre personnage dans un plan continu de 30 secondes
Importez une ou plusieurs photos nettes de votre personnage original (de face, de profil, en pied), puis :
30 seconds, 16:9, single continuous take. The young man from reference image 1, wearing the outfit shown in reference image 2, walks through a busy night market.
Keep his face, hair, and outfit consistent throughout.
Camera: handheld follow from behind, then moving to his side.
Style: warm lantern light, steam from food stalls, vibrant colors.
Sound: crowd chatter, sizzling food, a vendor calling out.
0–10s: He weaves through the crowd, looking around.
10–20s: He stops at a dumpling stall and points at the menu.
20–30s: He takes the first bite, burns his mouth, and laughs.
30. Produit à partir de vos photos
Importez des photos nettes du produit sous plusieurs angles, puis :
15 seconds, 1:1, single continuous take. The ceramic table lamp from reference images 1 to 3 sits on a bedside table in a calm bedroom at night.
Keep the lamp's shape, glaze color, and texture exactly as shown.
Camera: slow orbit around the lamp.
Style: dark room, the lamp as the main light source, warm and premium.
Sound: soft room tone, a gentle click when the lamp turns on.
0–5s: The room is dark.
5–15s: The lamp clicks on and warm light fills the room as the camera orbits.
Bonus pour les utilisateurs avancés : Dans les applications de ByteDance, Seedance 2.5 peut aussi utiliser des rendus d’argile comme références : des maquettes 3D grossières qui définissent la disposition, les poses des personnages et la trajectoire de la caméra. Si vous travaillez en 3D, ébauchez votre scène, exportez-la et utilisez-la comme référence avec votre prompt. Vérifiez que votre plateforme prend en charge ce mode avant de compter dessus.
Utiliser Seedance 2.5 sur AI Studio
Sur la page Seedance 2.5 d’AI Studio, vous trouverez :
- Durée : de 4 à 30 secondes, réglable à la seconde
- Résolution : 480p, 720p ou 1080p
- Formats : 16:9, 9:16, 1:1, 4:3, 3:4, 21:9 ou adaptatif
- Modes : texte vers vidéo ; image vers vidéo avec 1 ou 2 images ; références vers vidéo avec jusqu’à 30 images
- Audio : activé par défaut ; désactivez-le si vous prévoyez d’ajouter votre musique ou votre voix off
Le bouton de génération indique le coût en crédits avant de commencer. Les clips plus longs et les résolutions plus élevées coûtent davantage.
Quand préférer Seedance 2.0
Seedance 2.5 n’est pas toujours le meilleur choix. Pour un court clip image vers vidéo avec du son, le générateur vidéo Seedance 2.0 reste très performant. Dans le classement Artificial Analysis rapporté par GIGAZINE en août 2026, Seedance 2.0 est arrivé premier pour l’image vers vidéo avec audio. Il génère environ 15 secondes par passage, ce qui suffit pour la plupart des clips destinés aux réseaux sociaux.
Et si vous souhaitez voir votre prompt dans un tout autre modèle, le générateur vidéo IA d’AI Studio permet aussi de l’utiliser avec des modèles comme Kling 4.0, Gemini Omni et Happy Horse.
Sources
- ByteDance Seed, Introducing Seedance 2.5, 31 juillet 2026
- CineD, ByteDance Seedance 2.5 API goes live, juillet 2026
- GIGAZINE, MiniMax H3 video generation AI, 3 août 2026, pour le classement Artificial Analysis
- AI Studio, pages Seedance 2.5 et Seedance, consultées le 2 novembre 2026
Transformez votre idée en vidéo
Découvrez le générateur vidéo et commencez à créer.
Générer une vidéo


