Topview T1 : narration cinématographique et tarif 1080p prévisible
Génération en 1080p à un tarif publié de 2.2 crédits/s pour tous les membres, avec texte, image et entrée multimodale.
Quel modèle vidéo AI choisir ?
Seedance 2.5 et Topview T1 génèrent des clips vidéo AI jusqu’à 30 secondes, en sortie 1080p, avec un audio natif et jusqu’à 50 ressources de référence multimodales dans les flux Topview pris en charge. Topview T1 est conçu pour la narration cinématographique et publie un tarif de production 1080p de 2.2 crédits par seconde pour chaque membre. Seedance 2.5 ajoute des options de flux, dont un Draft Mode dédié en 480p, pour 3.7 crédits par seconde en sortie 1080p.
Regardez comment les modèles traitent le jeu dramatique, les scènes dialoguées, les clips musicaux, l’action fantastique et les effets visuels. Comparez le résultat créatif, les contraintes du flux et le devis réel en crédits avant de trancher.
Topview T1 : narration cinématographique et tarif 1080p prévisible
Génération en 1080p à un tarif publié de 2.2 crédits/s pour tous les membres, avec texte, image et entrée multimodale.
Seedance 2.5 : des contrôles de génération plus spécialisés
Sortie de 4 à 30 secondes, Draft Mode en 480p, conduite détaillée de la chronologie et flux de production assistés par la recherche lorsqu’ils sont pris en charge.
Les deux modèles prennent en charge un récit audiovisuel plus long et de grands ensembles de références. Les écarts concrets portent sur les paliers de sortie, certains modes de génération et la manière de payer chaque itération.
| Caractéristique | Topview T1 | Seedance 2.5 |
|---|---|---|
| Développeur / fournisseur | Premier modèle de génération vidéo de Topview | ByteDance |
| Durée maximale du clip | 30 secondes | 30 secondes |
| Réglages de durée | 4–30 s ; Smart Duration dans les modes API pris en charge | 4–30 s ; Smart Duration dans les modes API pris en charge |
| Résolution de sortie | 720p, 1080p | 480p, 720p, 1080p |
| Texte vers vidéo | Oui | Oui |
| Image vers vidéo | Oui | Oui |
| Références multiples | Oui | Oui |
| Total maximal de références | Jusqu’à 50 | Jusqu’à 50 |
| Références image | Jusqu’à 30 | Jusqu’à 30 |
| Références vidéo | Jusqu’à 10 | Jusqu’à 10 |
| Références audio | Jusqu’à 10 | Jusqu’à 10 |
| Audio natif | Disponible | Disponible |
| Format d’image | Paysage, portrait, carré et préréglages adaptatifs pris en charge | Paysage, portrait, carré et préréglages adaptatifs pris en charge |
| Draft Mode en 480p | Non pris en charge | Pris en charge |
| Prix en 1080p | Tarif publié de 2.2 crédits/s pour tous les membres | Dépend du flux, de l’offre et du devis en direct ; 3.7 crédits/s pour les utilisateurs normaux |
| Bon point de départ | Budget de production en 1080p et création menée par le récit | Contrôles spécialisés et flux qui commencent par un brouillon |
Les spécifications reflètent les flux Topview publiés. Elles ne garantissent pas que chaque option apparaisse dans chaque compte. Le nombre de références dépend du type de fichier, de la durée et des restrictions de mode.
Vérifiez dans le générateur les choix en cours avant d’envoyer.
Les deux peuvent produire une scène audiovisuelle complète de 30 secondes. Choisissez selon la contrainte de production qui pèse le plus : le prix d’une prise de plus, le flux d’entrée ou le degré de contrôle dont vous avez besoin.
Topview publie un tarif de 2.2 crédits/s en 1080p pour T1 sur les offres membres. Estimez le nombre de plans, la durée du clip et les reprises avant la production, puis confirmez le tarif affiché sur votre compte.
Explorez d’autres interprétations, d’autres angles et d’autres accents émotionnels sans vous enfermer dans une seule lecture. Topview T1 vise les courts drames, les scènes cinématographiques, l’animation et les clips où plusieurs prises peuvent être utiles.
Partez d’un prompt, d’une image de première image ou de ressources multimodales prises en charge. T1 accepte jusqu’à 50 éléments de référence dans les limites documentées d’entrée et de durée.
Générez la scène avec l’image et le son natif ensemble. Vérifiez les voix, les effets, le placement de la musique et le mouvement des lèvres dans l’exportation d’origine avant de l’intégrer à un projet fini.
Utilisez le Draft Mode pris en charge pour explorer une direction avant de valider un rendu final en plus haute résolution. Une vidéo 480p ordinaire et une tâche en Draft Mode 480p ne sont pas forcément facturées de la même manière.
Découpez une commande de 30 secondes en intervalles prévus pour les actions, le dialogue, le mouvement de caméra et le son. Vérifiez si la sortie finale suit cette séquence, surtout dans les scènes très physiques.
Utilisez la recherche internet uniquement lorsque le flux de références Topview que vous avez choisi la prend en charge.
Explorez dans Topview les outils de storyboard, de suite ou de montage par segments liés à Seedance. Avant de vous y fier pour le projet, distinguez les contrôles qui appartiennent au modèle de ceux qui appartiennent à la plateforme Topview.
Parcourez huit groupes de vidéos importées, du drame au suspense, en passant par la musique, l’action fantastique et les effets visuels. Chaque scénario isole un défi : jeu, continuité, identité, mouvement, son ou lecture de l’espace.
Caractéristiques du clip importé : 30s · 16:9 · 720p · 24fps
Une élève affronte le harcèlement dans une salle de classe ensoleillée. La scène passe de l’intimidation de groupe et des gros plans tendus au déplacement dans la classe, puis à un moment de résistance.
Topview T1
Seedance 2.5
Un court drame tient lorsque l’action et l’émotion restent lisibles. Un visage stable et un jeu expressif comptent davantage qu’une seule image séduisante.
Seedance 2.5 rend la fin parlée plus lisible. Les deux clips gardent la protagoniste en uniforme d’été, passent du groupe autour de son bureau à une bagarre en classe et montrent un coup de pied en l’air, les autres élèves au sol. Son visage reste reconnaissable dans les deux. Seedance conserve le gros plan agacé de face et termine sur un regard direct, bouche ouverte : l’avertissement se lit comme du jeu. Topview T1 tient aussi son identité et le coup de pied, mais le gros plan tardif est un profil caché par les cheveux, si bien que la dernière réplique se voit moins bien. Pendant la bagarre, Topview T1 garde un mouvement plus continu et moins raide. Choisissez Seedance 2.5 lorsque le tournant émotionnel et la fin parlée doivent rester lisibles.

CHARACTER REFERENCE — <<<Image1>>>. Cool and elegant Korean high school girl in a summer school uniform, realistic natural skin, no beautifying filters, and a consistent face throughout. Quiet and patient but strong boundaries. Her emotion evolves from quiet tolerance to cold annoyance, a controlled outburst and a tired, disdainful aftermath. high school girl, summer school uniform, natural skin texture, no filters, consistent facial features throughout. Quiet, avoids trouble, extremely patient but with strong boundaries. Emotional shift: Zen-like patience → thoroughly annoyed by repeated harassment → calm outburst → cold disdain. Post-fight: slightly messy hair, slight panting, realistic expressions. 00:00‒00:06 | Malicious Harassment (Suppression): Quiet afternoon classroom, female lead by the window. Four bullies provoke her, bumping her desk, pushing her, throwing stationery. She avoids and stays silent, wanting peace. Wide-angle handheld camera with slight shake, oppressive atmosphere. 00:06‒00:12 | Patience Exhausted (Turning Point): Slow zoom to facial close-up. Bullies escalate. Her patience vanishes, turning into cold annoyance. She stands up straight. Background blurs, silence except for her steady breathing, preparing to end the harassment. 00:12‒00:19 | Clean Counterattack (Main Fight): Girls push her. She blocks calmly, dodges quickly. Elbow strikes, precise low kicks. Realistic, clean movements. 360° wrap-around tracking shot with camera shake and motion blur. Three bullies fall, desks slide, scene reverses. 00:19‒00:24 | Mid-air Finisher (Climax Slow-mo): Last girl rushes. Ultra-low angle shot. Lead jumps, 35% slow motion. Uniform skirt and hair flow naturally. 360-degree mid-air roundhouse kick to the chest. Opponent hits the wall. Normal frame rate upon landing, clean and powerful. All four are down. 00:24‒00:27 | Silent Aftermath: Classroom in mess. Lead stands in center, panting slightly, cold gaze, tired and detached, calm and restrained aura. 00:27‒00:30 | Dialogue: Close-up. She looks at the camera, cold and annoyed tone. Korean line: "제발 귀찮게 하지 마." (Please stop bothering me.) Natural, realistic warning. No subtitles.
Caractéristiques du clip importé : 30s · 16:9 · 720p · 24fps
La nuit, sous une pluie forte, une voiture s’approche d’une laverie en bord de rue. À l’intérieur, un homme et une femme interagissent autour d’un comptoir, et la scène alterne l’extérieur, les détails d’atmosphère et les jeux en gros plan.
Topview T1
Seedance 2.5
Les intérieurs de nuit révèlent les failles de lumière, de continuité et de jeu fin. Ces détails portent le suspense et le récit filmé.
Caractéristiques du clip importé : 30s · 720p · 24fps
Un clip coloré de groupe féminin traverse une salle d’enregistrement rose, des couloirs et des moments de performance collective. On y voit une interprète aux cheveux roses, des tenues coordonnées et une lumière de studio vive.
Topview T1
Seedance 2.5
Un clip réunit identité, continuité entre scènes, chorégraphie et placement musical. Il rappelle aussi qu’il faut maîtriser l’exportation avant toute affirmation sur la qualité.
Caractéristiques du clip importé : 30s · 16:9 · 720p · 24fps
Deux interprètes évoluent sur un fond de studio jaune vif, avec batterie, amplificateurs, gestes énergiques et changements de position de caméra. La séquence alterne le corps entier, la percussion et l’interaction.
Topview T1
Seedance 2.5
Une performance musicale expose les faiblesses des mains, du rapport aux accessoires, de l’alignement sur le tempo et de la relation entre les personnes.
Caractéristiques du clip importé : 30s · 16:9 · 720p · 24fps
Des guerriers en armure affrontent d’énormes créatures fantastiques sur un champ de bataille enneigé. La scène mêle une formation, le froid, des effets magiques et une action de grande échelle.
Topview T1
Seedance 2.5
Une scène fantastique peut sembler spectaculaire tout en devenant confuse dans l’espace. Cette comparaison regarde si l’action se lit, si l’échelle tient et s’il y a une causalité.
Caractéristiques du clip importé : 30s · 16:9 · 1080p · 30fps
Une jeune femme affronte dans des rues de ville une créature géante aux yeux rouges, proche d’une araignée. La séquence comprend de la lumière sur les mains, des membres en mouvement, des véhicules, le mouvement du décor et de forts changements de distance de caméra.
Topview T1
Seedance 2.5
C’est la candidate la plus exigeante pour une comparaison d’effets aux mêmes dimensions de pixels à l’exportation. Elle sert à voir si le spectacle très mobile reste compréhensible sur une scène longue.
Caractéristiques du clip importé : 25s · 16:9 · 720p · 24fps
Une enquêtrice s’approche d’une scène de crime balisée dans une ville japonaise, avec un collègue plus âgé en costume. La scène traverse des rues résidentielles, des passants, des véhicules de police et le travail médico-légal.
Topview T1
Seedance 2.5
Un drame criminel réaliste repose sur un jeu retenu et un arrière-plan crédible. Le défi est de tenir ce réel, pas seulement d’augmenter les effets.
Caractéristiques du clip importé : 30s · 16:9 · 720p · 24fps
Une scène d’Halloween ludique se déroule dans un couloir peu éclairé, avec des citrouilles lumineuses. Un fantôme sous un drap apparaît lorsque des enfants costumés entrent et réagissent, et la surprise comique monte.
Topview T1
Seedance 2.5
La comédie et les courts métrages inquiétants exigent des réactions bien placées, des positions claires et une mise en scène stable. Cette comparaison vérifie si le modèle mène une histoire simple jusqu’au bout sans perdre ce dispositif visuel.
Nous regardons si chaque résultat sert la tâche créative d’origine, et non quelle image paraît plus dramatique.
Pour un test contrôlé, nous avons repris le même prompt d’origine, les mêmes sources, la durée visée, le format, la résolution et le réglage sonore partout où les deux modèles les prennent en charge.
Nous avons noté, pour chaque sortie, la durée réelle, les dimensions en pixels, la cadence, la piste audio, l’identifiant d’envoi et le modèle choisi. Ces propriétés techniques ne remplacent pas un jugement sur la qualité du modèle.
Nous avons utilisé une grille fixée à l’avance pour l’adhérence au prompt, la constance des personnages, le mouvement et la continuité de scène, l’alignement audiovisuel et l’usage pratique. Ce qui ne s’appliquait pas a été marqué comme sans objet.
Le dossier de test a gardé les tâches échouées, les sorties rejetées, les nouvelles tentatives et les vidéos réussies. Une tâche qui se termine sur le plan technique n’équivaut pas à une vidéo utilisable.
L’échantillon représentatif montré sur cette page utilise un nombre d’échantillons déclaré à l’avance et une notation à l’aveugle.
Le temps médian de fin se mesure de l’envoi au résultat téléchargeable pour les tâches terminées. Le taux d’achèvement utilise toutes les tentatives envoyées. Le coût par vidéo utilisable inclut les crédits des tentatives qui n’ont pas été acceptées. Taux d’achèvement = générations terminées ÷ toutes les tâches envoyées. Taux de vidéo utilisable = sorties acceptées ÷ toutes les tâches envoyées. Coût par vidéo utilisable = tous les crédits facturés ÷ sorties acceptées.
Une vidéo forte ou faible ne prouve pas qu’un modèle se comporte toujours ainsi. Les écarts peuvent venir des prompts, des références, de la variation aléatoire, des réglages ou du montage.
Topview publie actuellement un tarif T1 en 1080p de 2.2 crédits par seconde pour chaque membre, soit $0.22/s.
Les tarifs de Seedance 2.5 varient selon l’abonnement, le flux et les remises ; votre devis en direct peut montrer un écart plus petit ou différent.
Un projet peut comporter plusieurs plans et plusieurs essais par plan. Utilisez le tarif visible sur votre compte pour estimer le travail.
Crédits estimés = nombre de plans × générations par plan × secondes par génération × crédits par seconde.
Exemple de budget T1 :
Total estimé : 660 crédits pour 300 secondes générées.
Ce total est le métrage généré, pas la durée du montage fini. Le coût réel peut changer si vous utilisez un autre flux, une promotion ou un avantage d’abonnement.
Consulter les prix TopviewPour une scène de dialogue ou un court drame, plusieurs interprétations peuvent être nécessaires avant que le montage tienne. Le tarif membre 1080p publié de T1 rend plus simple l’estimation du coût de ces essais.
S’il vous faut seulement jusqu’à 30 secondes, une sortie 1080p, un audio natif et un ensemble de références dans la limite de 50 ressources, les deux modèles offrent cette base. Testez la qualité et le coût au lieu de supposer qu’une autre spécification est nécessaire.
Seedance 2.5 propose un Draft Mode documenté que T1 n’offre pas. Cela peut servir lorsque le premier objectif est de valider un plan avant une génération finale en 1080p.
La génération de références assistée par la recherche internet et certains contrôles créatifs ou de montage propres au modèle ne sont pas également disponibles sur T1 à ce jour. Vérifiez le mode exact avant de planifier le processus.
Générateur vidéo AI Topview T1
Parcourez la vitrine cinématographique de T1, le texte vers vidéo, l’image vers vidéo, les flux à plusieurs références et le prix 1080p actuel.
Explorer Topview T1
Générateur vidéo AI Seedance 2.5
Passez en revue les références prises en charge, les contrôles créatifs en code temporel, les réglages de sortie et les exemples vidéo.
Explorer Seedance 2.5
Comment utiliser Topview T1
Suivez un parcours pas à pas pour préparer un prompt, choisir les réglages et générer une première vidéo.
Lire le guide Topview T1
Comment utiliser Seedance 2.5
Suivez les étapes pour monter un prompt, fixer les réglages et obtenir une première vidéo.
Lire le tutoriel Seedance 2.5
Prix de l’API pour Topview T1 et Seedance 2.5
Consultez les tarifs en vigueur de la plateforme pour la génération par API avant d’estimer une intégration de production.
Voir les prix de l’API
Seedance 2.5 et Seedance 2.0 : comparaison
Comparez les versions du modèle et regardez d’autres exemples vidéo appariés.
Comparer les versions de Seedance
Topview T1 et Seedance 2.5
De meilleures histoires demandent le bon flux
Choisissez Topview T1 lorsque le coût de production 1080p prévisible et des prises créatives supplémentaires comptent. Choisissez Seedance 2.5 lorsque votre processus dépend du Draft Mode ou de contrôles de génération plus spécialisés.
Topview T1 correspond mieux à cet essai de jeu. Les deux clips ouvrent sur une laverie sous la pluie et une voiture, coupent sur un mesureur de signal et finissent sur la berline dans la rue mouillée. L’intérieur de T1 est la scène la plus nette : un plan à deux assis, des mains et des papiers sur le comptoir, un phare qui traverse le visage de la femme, puis le duo au travail entre les machines. Seedance 2.5 tient la pluie et le mesureur, et sa fin montre le chariot de linge avec la voiture derrière, ce qui énonce le leurre de façon plus littérale. Le milieu, lui, ressemble davantage à du pliage de linge qu’à un échange tendu, et un plan plus tard est surtout la voiture et un faisceau de lumière. Choisissez T1 lorsque l’intérieur de nuit et les deux interprétations doivent porter la scène. Choisissez Seedance 2.5 lorsque le chariot et la voiture doivent se lire dans un seul plan.
Create a 30 second cinematic short film titled The Suit Was Still Listening, presented in 16:9 with synchronized dialogue, environmental sound, and original music. The film must feel completely live action, physically grounded, and captured on location, never glossy or synthetic. Visual language: a rain soaked neighborhood laundromat at midnight, warm amber fluorescent tubes inside, cold blue street light and moving car reflections outside, wet pavement, fogged windows, chipped enamel machines, baskets, detergent boxes, loose receipts, realistic skin texture, natural eye moisture, subtle fabric lint, imperfect hair, restrained film grain, gentle halation, lifted shadows, rich but believable color. Wardrobe creates bold editorial color blocking: Mara wears a cobalt trench over a cream blouse, Noah wears a rust knit polo, the teenage attendant wears faded green workwear, and an elderly customer wears a pale gray suit and dark hat. Use expressive medium close shots, tactile inserts, occasional low angles, clean profile compositions, and brief handheld movement that feels operated by a human. Preserve consistent faces, clothing, props, screen direction, reflections, hand anatomy, and spatial continuity. Use realistic speech timing, breathing, blinking, eye focus, and precise lip synchronization. 0 to 3 seconds. Exterior wide shot through falling rain. A lonely laundromat glows on a dark corner. A black sedan turns slowly into the street. Camera makes a subtle forward creep. Music begins with muted upright bass, brushed snare, a low analog pulse, and the rhythmic churn of washers. 3 to 6 seconds. Interior macro shot through a round washer door. A red scarf turns through soap and water. Cut to Mara opening a dryer. A tiny black transmitter falls from the lining of a cream suit and strikes the tile with a sharp metallic click. The score briefly drops out. 6 to 10 seconds. Low medium two shot. Noah freezes beside a folding table. He says quietly, You said clean the suit. Mara studies the transmitter in her palm and replies, I did. You hid a heartbeat in the lining. Keep the delivery controlled and intimate, not theatrical. 10 to 14 seconds. Tight insert on the teenage attendant checking an old signal meter beside the change machine. A green needle jumps. He looks toward the fogged front window and says, It is still transmitting. Camera racks focus from the meter to white headlights sliding across the glass. 14 to 18 seconds. The sedan stops outside. The elderly customer continues folding a shirt without looking up. He says, Then he is already here. Hold on Mara as the moving headlights carve across her face. Let silence, rain, and machine motors carry the tension. 18 to 22 seconds. Noah whispers, Back door. The old man calmly turns the deadbolt on the back exit and answers, Too obvious. Mara notices a delivery rider outside pushing a rolling laundry cart past the window. Use a fast sequence of close shots: her eyes, the transmitter, the cart wheel, the sedan mirror. 22 to 27 seconds. Mara crosses naturally behind a row of machines, slips the transmitter into a sealed laundry bag on the moving cart through an open service hatch, and returns without drawing attention. Noah asks, What did you do. Mara watches the cart continue down the wet street and says, Gave him a cleaner story. 27 to 30 seconds. Exterior telephoto view. The black sedan pulls away and follows the cart. Cut back inside to a symmetrical medium shot of the four characters framed by spinning washer doors. The red scarf circles behind Mara like a slow warning light. She sits opposite Noah and says, Now tell me who owns the suit. End on the transmitter signal fading from the attendant meter as the final bass note lands. Sound design must include rain on glass, distant tires on wet road, washer motors, dryer buttons, fabric movement, the transmitter click, door lock, fluorescent hum, and soft room reflections around every voice. Dialogue must remain clean and naturally mixed above the music. Avoid visual distortion, artificial camera acceleration, floating objects, excessive shallow focus, oversharpening, plastic skin, extra fingers, unreadable signage, random background motion, or dreamlike effects.
Seedance 2.5 rend les quatre interprètes plus faciles à distinguer. Les deux clips gardent les quatre styles : cheveux blonds et tenue rose, chignon bleu, longs cheveux sombres en crème, cheveux rose lavande en bleu. Le plan de groupe le plus net de Topview T1 est une formation synchronisée de quatre personnes autour d’un anneau argenté. Seedance 2.5 montre la membre aux cheveux roses avec un téléphone à clapet et la bouche ouverte, puis une fin en quatre personnes décalées, et non une formation verrouillée. Choisissez Seedance 2.5 lorsqu’un solo et un accessoire doivent survivre à la coupe. Choisissez Topview T1 lorsque le plan qui doit se lire est la formation du refrain.
Référence audio





Create an original 30-second, 16:9 live-action Y2K sweet-and-edgy K-pop girl-group music video, with four adult female performers. Use Audio 4 (<<<aud_4>>>) as the single authoritative reference for the song, lyrics, vocal delivery, beats, and edit timing. Preserve the four alternating solo verses, four short group-response moments, and the final full-group chorus. Solo lyrics must be accurately lip-synced; group-response shots can feature playful interactions without forcing every member to mouth each word. Keep the track, bass, drums, and stage lighting at full intensity until 28 seconds. Only begin a smooth musical and visual fade during 28‒30s. The final five seconds must feature four distinct changing compositions― not a static five-second push-in. CHARACTER REFERENCES ― KEEP ALL FOUR IDENTITIES DISTINCT: Member 1 (Image 25): Light blonde high ponytail, pink collegiate Y2K outfit; preserve her face, proportions, hairstyle, outfit, and accessories. Member 2 (Image 20): Blue updo, powder-blue collegiate look; preserve the separate face and costume identity. Member 3 (Image 15): Long dark loose hair, cream collegiate outfit; preserve the separate face and costume identity. Member 4 (Image 24): Long pink-lavender hair, light-blue sporty collegiate outfit; preserve her face, silhouette, and accessories. Image 2: Group relationship, pastel pink/blue/cream styling, overall Y2K sweet-cool visual direction. Never blend faces, swap outfits, clone members, or introduce additional performers. 0‒3s | MEMBER 1 SOLO ― PINK CLASSROOM FANTASY SET. Cut directly to the blonde singer leaning against a transparent desk. Soft gradient pink wall, venetian-blind shadows, cream storage modules, floating abstract panels. She looks sideways, turns to camera, and lip-syncs: “Bell rings, cherry gloss, I’m out the door.” Playful eyes, natural turn and small hand gesture; no full choreography. 3‒5s | GROUP RESPONSE ― DESK PLAY. On “Hey! We’re ready for more,” Member 1 swipes a translucent card across the lens to transition. Four members gather around a clear acrylic table, swap bright notes, playfully bump shoulders, gesture to camera, and laugh. Keep the original vocals; line-by-line lip-sync is unnecessary in this cutaway. 5‒8s | MEMBER 2 SOLO ― ICE-BLUE AND CREAM MODULAR SET. A reflection along a mirror bench wipes to Member 2. The scene is predominantly cream with touches of low-saturation ice blue, fine silver framing, clear acrylic foreground, lilac frosted-glass screen, blush detail, and distinct foreground/midground/background layers. Never flood the frame with flat blue light. Member 2 steps forward with restrained cool confidence and accurately lip-syncs: “Flip-phone flash, I’m stealing the light.” Subtle push-in; focus remains on her eyes. 8‒10s | GROUP RESPONSE ― TRANSPARENT ROLLING CHAIR. She pushes a reflective panel across the lens, revealing the pastel lounge. Member 1 sits on a transparent rolling chair, Member 4 gently pushes it, and Members 2 and 3 jog beside them laughing. During “Oh! We own the night,” use a lateral tracking shot and spontaneous friend-group interaction; no required individual lip-sync. 10‒13s | MEMBER 3 SOLO ― CREAM MEDIA-CLASSROOM FANTASY. The rolling chair wipes into a soft cream set with modular shelving, translucent divider, faint pink blind-shaped light, thin acrylic console and tiny ice-blue indicators. Member 3 taps the console, glowing strips illuminate, and she raises her eyes toward the moving camera while lip-syncing: “Pastel dreams in the window shine.” Gentle breathing, turning and eye contact rather than dancing. 13‒15s | GROUP RESPONSE ― DIGITAL CAMERA FLASH. The console light blooms into a digital-camera flash transition. Four performers crowd together in a clean pink-and-cream photo corner, straighten each other's hair clips, lean shoulders together, and change funny expressions on the second flash. Retain “Yeah! This moment is mine” on the soundtrack without mandatory detailed lip-sync. 15‒18s | MEMBER 4 SOLO ― LAVENDER COMMUNICATION ROOM. The flash clears onto an original lilac modular studio with silver slim framing, frosted acrylic, cream low cabinets, soft horizontal lighting, and a transparent flip-phone prop. Member 4 lowers the receiver and approaches camera half a step, confidently lip-syncing: “Turn up the hall, let the whole school know.” Natural hair movement, open gesture, no complex dance. 18‒20s | GROUP RESPONSE ― ENTERING THE MAIN ARENA. During “One, two, three―let’s go!” Member 4 slides open a translucent divider. The other three peek out, link hands, and run forward into a glowing doorway. Their movement carries seamlessly into the final arena. 20‒23s | FINAL CHORUS ― UPGRADED STAGE. Reveal an original “After School Club Arena”: translucent curved rear wall, abstract locker-shaped arches at either side, floating silver elliptical structure, cream-and-light-wood reflective sports floor, pink-and-blue illuminated floor tracks, transparent raised platforms. All four perform synchronized full-body choreography and sing: “After school electric, e-e-electric.” The camera glides forward through the glowing doorway. 23‒25s | CHORUS PROGRESSION ― DIAMOND FORMATION. The members move from a horizontal line into a diamond, singing “Candy-coated confidence, magnetic.” Platform light strips respond like sound waves. The camera tracks from knee-height framing out to all four performers. Keep drums, bass, and vocals vivid. 25‒27s | CLIMAX A ― SIDEWAYS ARCUATE TRACK. On “Four hearts, one beat, lighting up the room,” the camera sweeps from stage front-left to front-right with foreground parallax from transparent platforms. The dancers interlace arm gestures, cross positions in two steps, half-turn, and open into a wide diagonal formation. Alternate full-body and knees-up framing, preserving all four identities. Music and lighting remain at their highest energy. 27‒28s | CLIMAX B ― LOW-ANGLE IMPACT. Hard cut to a slightly low-angle group shot, push in briefly then arc around the outer side. They clearly sing “Hands up together―watch us go boom!” together and land a precisely synchronized arm-raise, turn, and step on the final musical impact. The silver frame flashes from ice blue to bubblegum pink and floor strips ignite in sequence. No early musical fade before 28s. 28‒29s | FADE A ― PAIRED MEDIUM CLOSE-UPS. The lyrics have ended; retain wordless group harmony, synthesizer sustain, final drum tail, bass resonance, and reverberation while the volume steadily begins its two-second fade. Show a glance-to-camera from Members 1 and 2, wipe via hair and shoulder motion, then show Members 3 and 4 trading subtle expressions, each with a brief semi-orbit. Background lights dim in layers but faces stay legible. 29‒30s | FADE B ― FOUR-MEMBER FINAL HERO FRAME. Cut to a wide slightly low-angle group composition with an asymmetric staggered arrangement, two women at the front of the transparent platforms and two on a higher level. Keep only natural breathing, settling hair, and slight eye movement. The camera floats sideways and slowly pulls back, settling during the final half-second. Architecture, silver ring, reflective floor and pink-blue tracks fade softly with the last instrumental decay. No abrupt freeze, harsh cut, or unfinished ending. TRANSITIONS, STYLE AND QUALITY: Link scenes through practical on-camera swipes: translucent card, mirror wipe, rolling chair occlusion, camera flashes, sliding acrylic divider, and illuminated doorway. All solo studios share cream modular architecture, blind-shaped soft shadows, pale acrylic and subtle chrome, but each retains unique accents. Premium live-action music-video studio production, early-2000s fashion-magazine energy, bubblegum pink, ice blue, cream white, muted lilac, selective silver and black. Campus elements must be abstract and editorial rather than a realistic school. Real human skin, accurate facial identity and lip-sync, energetic but readable cinematography. No microphones, in-ear monitors, subtitles, on-screen text, extra people, or brand logos. PROTOTYPE ASSET NOTE: The referenced Audio 4 and Images 2, 15, 20, 24, 25 are source-material IDs from the production brief; they are not bundled in this preview HTML.
Seul Topview T1 suit entièrement les actions de la vidéo de référence. Les deux clips utilisent le studio jaune, et les costumes restent stables dans les deux : cheveux rouges, corset noir, chemise à carreaux, casquette à l’envers et pantalon crème. Seedance 2.5 est un peu plus net sur la danse bras levés et sur la veste noire pendant qu’elle est en l’air, mais il ne colle pas à l’action de référence mouvement par mouvement comme le fait T1. Choisissez Topview T1 lorsque la performance doit correspondre à la vidéo de référence mouvement par mouvement.


Référence vidéo
【TASK: Frame-by-frame remake of the reference video ― replace only the specified characters and studio color】 Use <<<Video1>>> as the sole reference for action, performance, character interaction, camera move, framing, edit rhythm, and timeline order. Fully replace the original male with the man in <<<Image2>>>, and the original female with the woman in <<<Image1>>>. Change the pink studio walls and floor to a bright, saturated, clean yellow. This is reference-video-driven character and set replacement ― not a new story, not a redesigned dance, not freestyle to the song. Final length ~26.67s, 16:9, 1920×1080, 30fps. Keep real action speed from the source ― no slow-mo, no stretch, no compress to 15s. Timecodes below are alignment aids; cuts, action starts/ends, and paths follow <<<Video1>>>. Only <<<Image2>>> male lead and <<<Image1>>> female lead ― no third person, dancers, crowd, or clones. 【1. Source roles & priority】 Video1: camera, staging, body action, expression timing, mouth, interaction, props, cuts, aspect inserts, light 2D doodle FX only. Do not inherit original actors’ faces, hair, beards, hats, or clothes. <<<Image2>>> locks: male face, skin, short curls, backward black baseball cap, build, red-white check shirt, cream wide trousers, chunky shoes, right-arm tattoo, wrist accessories. <<<Image1>>> locks: female face, long cherry-red hair with face-framing micro-braids, build, white baby tee under black lace-up corset vest, black pleated mini skirt, silver chain choker, flesh-tone headset mic. On conflict: looks/wardrobe from Image1/Image2; action/camera from Video1. Only micro-adapt kneel/sit/jacket-toss for the new clothes ― do not rewrite whole beats. 【2. Male consistency <<<Image2>>>】 Same adult man: dark skin, lean athletic, short dense black curls (tapered sides), clean jaw, faint upper-lip shadow only. Always wears a plain black baseball cap backwards (visor at the back). No glasses. Not the original actor’s face. Wardrobe always: oversized red-and-white windowpane short-sleeve shirt, extremely baggy cream/beige wide-leg trousers, grey/off-white chunky platform shoes. Multi-strand white beaded bracelets on right wrist; silver metal watch on left. Large black tattoo on right forearm wrist-to-below-elbow, inner and outer. Fixed to that same right arm ― never swap sides, never mirror both arms, never copy to upper arm/hand/clothes. Perform drums, nods, leans, dance, claps, floor games, cup sips, side-lie, and lip-sync as in Video1. Expressions change with performance. Plain black sunglasses and star party glasses are temporary props; after removal, no extra sunglasses remain on his face. No prop pop-in/out mid continuous shot. 【3. Female consistency <<<Image1>>>】 Same adult woman: East Asian, pale skin, slim, long cherry/crimson-red hair slightly wind-moved with two thin face-framing braids, sharp winged liner, blurred red lips, cool girl-crush look. Any facial mole stays fixed ― no mirror flip, no vanish in close-ups. Wardrobe always: white short-sleeve baby tee under black lace-up corset vest with silver grommets/buckles, black pleated mini skirt with grommet belt and side silver chain, silver chain choker, small hoops, multiple silver rings; flesh-tone headset mic near mouth. Never gold hair, green cami, burgundy floral skirt, brown boots, or other original looks. Keep playful energy: sing to camera, reach, head toss, shoulder lean, remove glasses, dance, clap, point laughing, cup sip, prop toss. Hair must be her own long red hair with natural inertia. Mic stays on face; on kneel/sit the mini skirt covers naturally ― not a long dress or pants. 【4. Set】 Keep open studio: seamless wall meets floor; speakers/gear left-right; drum kit scale/placement in drum beats; later dance/floor beats use matching gear layout (don’t force kit into dance zone). Replace all pink walls/floor with bright saturated clean yellow (real softbox falloff ― not neon, not orange-pink residue). Solid studio floor with contact shadows. Speakers black/dark grey; metal stands real. Recolor environment only ― no full-frame yellow grade on skin/clothes. Kick drum face clean black, no art/text. No logos, charts, coins, finance UI. 【5. Camera & edit】 Match playful retro music-video studio feel. Keep mid frontals, high angles, slight Dutch, close wide, mouth ECU, two-shots, fast cuts, short punches. No new slow-mo, long orbits, drone, audience, or redesigned oners. Real faces/fabric; light grain and motion blur. Keep letterboxed inserts and slight tilt. Output canvas 16:9. Strip platform watermarks/UI; no new titles/end cards. 【6. Shot timeline】 (cuts/action from Video1) Shot 01|00.00‒01.10|She sings foreground, he drums background Lowish frontal medium-wide. She kneels left FG with wired mic, lip-sync; red hair, white tee + black corset vest, black mini clear. He at kit rear-right: backward black cap, red-white check, cream trousers, drumming on beat; right-arm tattoo when visible. Clean black kick face. Shot 02|01.10‒04.30|High angle, she reaches to camera Free hand reaches into lens (strong FG perspective), then retract, point, head sway, lean; red hair moves. Camera rises/pulls to show more yellow floor and kit. He keeps drumming BG. Optional tiny lightning/star doodles at source hits. Shot 03|04.30‒07.67|Letterboxed shoulder interaction Black bars, slight tilt. He left at drums with sticks; she right leans on his shoulder ― arm wrap, side bend, head whip (red hair swings) per source. He stays seated nodding/tapping. Shot 04|07.67‒10.80|She takes his sunglasses, wears them Same physical black sunglasses: she lifts off his face, stands, turns, light hair toss, puts them on herself, re-leans. He has no thin eyeglasses; after transfer he must not still wear an identical pair. Backward cap stays on. Shot 05|10.80‒12.30|Letterboxed high two-shot She in the taken sunglasses leans on him lip-syncing; he keeps stick rhythm, brief eyes-closed/smile/tilt. Faces stay Image2 / Image1 distinct. Shot 06|12.30‒13.93|Her mouth ECU Her own lips/teeth/skin from Image1; mouth follows source syllables. Soft warm-gold instead of pink wash ― don’t yellow-tint lips/teeth. Shot 07|13.93‒15.57|Full-body synced dance He left, she right, full body: backward cap + check shirt + cream trousers + platforms; red hair + white tee/black vest + black pleated mini. Beat-for-beat elbows, shoulder pops, leans, weight shifts ― no mirror swap, no random waving. Shot 08|15.57‒16.37|Her close, him BG She FG head tilts/nods lip-sync; red hair + vest clear. He keeps dancing BG. Shot 09|16.37‒16.97|His close, her BG He FG lean, lip-sync, short chest gestures; backward cap, curls, check shirt locked. Tattoo only when angle shows it. Shot 10|16.97‒17.37|Side clap transition Side-turn steps + chest claps on downbeats; optional tiny line/lightning doodles. Shot 11|17.37‒20.47|Face-to-face floor clap game He left facing right, she right facing left; center small cup. Full clap → reach → contact → retract → clasp cycles; no sticky fingers, pass-through, or extra arms. Skirt and trousers sit naturally. Shot 12|20.47‒22.10|Giant star glasses sing to camera She left, he right, oversized five-point star party glasses; nod/tilt/lean lip-sync. Lower faces + hair/cap still readable. Shot 13|22.10‒23.40|Floor game, she points laughing She points at him, leans back laughing; he reacts head-down, hand up, cup, lean-back ― cause/effect intact. Shot 14|23.40‒23.80|Punch-in: she sips cup Fast push + blur; cup/lips contact real; red hair swings back. Don’t stretch to slow-mo. Shot 15|23.80‒24.30|He sips cup CU Same sip/lean rhythm; always Image2 face + backward cap + check shirt. Shot 16|24.30‒25.80|He FG side-lie; she BG swings/throws jacket He lower-left side-lying in check shirt, cream trousers, platforms, backward cap. She mid-rear swings one separate black jacket prop (not clothing she is wearing); she stays in white tee + black corset vest + black mini. Jacket path: side swing → toss up → flip → skim FG/exit per source. Never remove the skirt or switch to white top/black shorts. Shot 17|25.80‒26.67|Two-shot look/laugh end He left, she right, look at each other; side head, laugh up, brief look down, light sway. Keep Image2 / Image1 looks. No freeze portrait, hug, kiss, titles, or end card. End on source final frame. 【7. Sound & lips】Keep original track, tempo, vocal order, length. Lips follow syllables; drum hits and claps sync to contact. Don’t drown music with SFX. Same timeline ― don’t retimed the song. 【8. Hard avoids】No new plot/dance; no meeting/award/stage-crowd rewrite; no face-swap that keeps blonde/cap/leather of originals; no matching yellow outfits; no changing check shirt/cream pants or red-hair corset/black mini; no identity drift / tattoo side-swap; no prop teleport / sunglasses clone; no pink residue / full yellow grade; no subtitles, finance UI, watermarks; no slow-mo or stretched shots. Final goal: same shots, same actions, same interactions, same edit ― only the cast fully replaced by <<<Image2>>> male (with backward black cap) and <<<Image1>>> female (cherry-red hair + black corset look), and pink studio → clean yellow studio.
Topview T1 rend l’échelle plus facile à lire. Il ouvre dans la tente cramoisie, avec les bannières au soleil et le capitaine, puis un géant au-dessus d’un mur de boucliers, un balayage violet sur les géants liés, une charge dans la neige, des ours polaires qui restent immenses à côté des soldats, et un dernier plan où la tête d’un ours remplit le cadre près du visage du capitaine. Seedance 2.5 ouvre dans la forêt, et non dans la tente. Son géant tient moins de place dans le cadre, la charge dans la neige garde encore un géant derrière les coureurs, et le moment des ours inclut toujours un géant de pierre. Le gros plan final du capitaine est le visage le plus net, avec les ours derrière lui. Choisissez Topview T1 lorsque la taille des créatures et le passage de la tente à la neige doivent rester évidents. Choisissez Seedance 2.5 lorsque le visage final du capitaine doit être l’image la plus nette.
A 30-second ultra-realistic 4K cinematic original war short, ultra-high saturation modern film look (crisp vivid film grain, not vintage). Fully original fantasy world ― invent all designs, no resemblance to any existing film, game, book, or myth franchise. No copyrighted characters, creatures, armor styles, or logos. EXACT ORDER: 0‒4s ― THE HOOK (50mm slow push → 85mm): A mixed medieval-fantasy army at a dark pine forest edge, holding deep crimson hand-painted cloth banners with simple geometric sun-disk emblems (original, non-trademark). A child in torn clothes flashes past crying for one second, then exits forever. A scarred captain in custom riveted iron plate and a worn crimson cloak whispers English only: “What is that?” Soldiers exchange uneasy looks. Oppressive quiet, shallow DOF, out-of-focus banner cloth in foreground bokeh. 4‒9s ― THE EMERGENCE (28mm low-angle wide): Thin mist parts. A line of original colossal humanoid war-constructs steps out ― rough basalt-skin bodies, rope-bound wooden plating, bone charms, unique silhouette unlike any known giant IP. They tower over the banners. Captain bellows: “RUN!” Rack focus from a spear tip in foreground to a massive stone foot. 9‒13s ― SCAN 1 / FOREST→SNOW: Soft freeze stutter. Very low-key pale purple local scan sweeps trees only ― thin, restrained, not flashy. Forest hard-replaces to open snowfield; same army and constructs, continuity held. 13‒22s ― RUNNING WAR (35mm steadicam + 85mm ECU): Soldiers flee while fighting. Constructs smash fleeing lines; spears snap, men hurled into snow, crimson banners drag and fall. Captain briefly counters then keeps running. Tracking beside panicked soldiers, OTS strikes, ultra-close practical impact inserts (boot, shield, snow burst), whip pans, half-beat pauses. No magic energy beams. 22‒26s ― SCAN 2 / CONSTRUCTS→WHITE BEARS: Soft freeze. Pale purple local scan only across construct bodies; they hard-replace into same-scale giant wild polar bears (natural animals, no armored fantasy brands). Action unbroken: bears roar and smash, soldiers keep fleeing and fighting. 26‒30s ― AFTERSHOCK (28mm wide → 85mm punch-in): Chaotic snow war ― polar bears advancing, banners collapsing, soldiers scattering. No child. End on captain’s exhausted determined face racked into sharp focus, a bear silhouette soft behind him. No logos, no text, no watermark. CAMERA LOCK (key nodes only): 28mm low-angle scale, 35mm chase tracking, 50‒85mm faces, rack focus, OTS, shallow foreground bokeh. Anamorphic cinema feel, fine film grain, ultra-high saturation crimson accents, breath fog. Purple scans subtle and local only. No neon, no ruins, no kinetic FX, no franchise lookalikes. English dialogue only.
Topview T1 raconte le combat en temps séparés. Les deux clips gardent une femme aux cheveux bleus en robe vert citron, dans une rue mouillée, avec des taxis jaunes et une seule grande araignée. T1 passe du camion à l’araignée dessus, puis à son coup de poing avec de la lumière sur les mains, un gros plan de patte, l’impact, un taxi retourné, des yeux lumineux et un portrait final. Seedance 2.5 rend plus tôt évidente la taille de l’araignée par rapport à la rue, y compris la femme debout devant l’araignée et le taxi, mais le faisceau lumineux se lit comme venant de l’araignée, et la fin répète des gros plans d’approche similaires. Choisissez Topview T1 lorsque le coup de poing et les mains lumineuses doivent se lire seuls. Choisissez Seedance 2.5 lorsque l’échelle de la créature dans la rue compte davantage.

Create a 30-second, 16:9 photorealistic live-action action sequence on an overcast Manhattan street after rain. Wet asphalt reflects yellow taxis and gray buildings. [Subject] is the protagonist. Keep the protagonist’s identity and appearance consistent throughout, without specifying gender, facial features, hairstyle, or clothing. The monster is one giant reddish-brown spider with exactly eight clearly articulated legs and a cluster of glowing red eyes. Maintain a consistent street layout: the white box truck is up the street, an overturned yellow taxi is on the protagonist’s left, and a parked car is on the right. Keep the action on one side of the street axis so the direction of movement remains clear. 00:00‒00:03 ― Street and warning Wide establishing shot from a low, 24mm street-level camera tracking beside a shallow puddle. A yellow taxi crosses the foreground, briefly wiping the frame. As it clears, the white box truck comes into view up the street and jolts violently. Make a quick handheld push toward the truck. Original, wordless male humming begins quietly beneath traffic noise. 00:03‒00:07 ― Spider reveal Cut to a tight, low-angle shot of the truck’s rear doors buckling outward. The camera snaps back to a wider three-quarter view as the cargo box tears open and the spider emerges. Follow one leg striking the asphalt in a brief ground-level insert, then tilt up to reveal its full scale. It overturns a taxi; the camera operator instinctively stumbles backward as pedestrians run past the lens. Keep all eight legs visually distinct. 00:07‒00:11 ― Protagonist reveal Cut behind [Subject] in a medium over-the-shoulder shot, with the spider visible up the street. Arc the handheld camera around to a front medium close-up without crossing the action axis. [Subject] stops rather than running. Rack focus from the spider in the background to pink electrical arcs gathering around the protagonist’s fists. Finish with a quick close insert of electricity reflected in the wet road. 00:11‒00:15 ― First strike From a low side angle, track laterally with a spider leg as it thrusts toward [Subject]. Cut wider before the dodge so the complete movement is readable. [Subject] steps aside and punches the leg joint. Switch to a tight impact shot for a fraction of a second: the fist connects, pink electricity branches across the leg, and droplets jump from its surface. Use only a brief slow-motion accent at contact, then return to real speed. 00:15‒00:19 ― Leap and counterpunch Start with a wide shot that shows [Subject], the overturned taxi, and the spider in the same frame. A low tracking camera follows [Subject] running toward the taxi and using it to jump. Cut to a rising side angle that follows the jump’s arc; avoid an impossible midair camera orbit. [Subject] lands on the spider’s cephalothorax. A tight, slightly shaky overhead shot catches the downward punch, followed by a wider reaction shot as the spider recoils. 00:19‒00:23 ― Thrown into the car Hold a medium-wide shot long enough to show the spider’s physical movement throwing [Subject] toward the parked car on the right. Pan with the trajectory rather than cutting away. The car door dents on impact and its windows vibrate. Cut to a close, canted handheld shot beside the car: sound drops into muffled ringing, [Subject] grips the bodywork and rises, and pink light briefly appears in their eyes. Preserve the protagonist’s face and identity. 00:23‒00:27 ― Electrical blast Begin behind [Subject] in a low over-the-shoulder shot, framing both extended fists and the spider down the street. [Subject] releases one concentrated pink electrical blast. Cut to a wide frontal view of the spider so its resistance is visible: all eight legs dig into the wet asphalt before it slides backward several meters. Use a short close insert of sparks and water spraying from its feet. Return to the wide shot to show the full distance it has been pushed. Do not turn the blast into an explosion. 00:27‒00:30 ― Shock and silence As the electricity fades, follow [Subject] descending into the intersection with a gentle handheld tilt down. Move into a slow, restrained push-in from medium shot to close-up. [Subject] breathes hard and looks toward the distant spider, visibly shocked by what just happened. The music cuts out; only rain, electrical fizz, and a faint siren remain. [Subject] quietly says, “Stay there.” Hold for a beat with subtle handheld movement, then cut to black. Overall camera language: Immediate, physical handheld photography; grounded low angles that establish scale; clear wide shots before complex actions; close inserts only for decisive details. Use brief slow motion at the two major punches, never throughout the fight. Natural overcast light and cool gray streets contrast with vivid pink electricity. Give every jump, collision, and monster movement realistic weight. Maintain consistent protagonist identity, spider anatomy, vehicle positions, and screen direction across every cut. No on-screen text. Audio: Use original, wordless vocal textures and original ambient scoring only. Build the action with tire skids, metal impacts, footsteps, spider legs striking asphalt, rain, and electrical crackle. Do not use, quote, reference, or imitate any existing song, melody, recording, or artist. Negative prompt: cartoon, anime, plastic skin, blurry action, warped face, changing protagonist identity, extra human limbs, bad hands, spider with fewer or more than eight legs, fused or mushy legs, changing monster design, weightless impacts, rubbery vehicle damage, sunshine, impossible camera movement, confusing screen direction, gore, text, watermark.
Topview T1 rend plus facile à suivre le trajet des enquêteurs et leur place dans la scène. Les deux clips gardent la femme en blazer noir et l’homme plus âgé en costume, puis atteignent le ruban jaune, une voiture de police japonaise et des combinaisons médico-légales bleues. Le lettrage des deux voitures est un marquage de police correct. Sur la voiture de T1, le caractère 庁 est répété. T1 montre encore le trajet le plus clair : elle arrive par une ruelle résidentielle calme, passe la voiture et rejoint le marqueur entouré de ruban, avec les enquêteurs à genoux, si bien que la direction de la scène reste lisible. Seedance 2.5 ouvre déjà au milieu de la foule. Choisissez Topview T1 lorsque l’approche et l’organisation du lieu doivent rester claires.


Create a 25-second photorealistic Japanese crime-drama sequence in a quiet residential district of Tokyo, filmed in ONE UNBROKEN TAKE: no cuts, invisible cuts, time skips, location changes, or editorial transitions. Format 16:9 landscape, 30 fps, with high-end 4K-capture aesthetics and an observational TV-drama/documentary crime-scene feel. Restrained, natural and professional acting―no fashion-commercial posing, heroic introductions, or exaggerated shock. LOCATION AND TONE: A typical Tokyo apartment neighborhood at about 4 p.m. on a fully overcast workday. Narrow asphalt road between mid-rise apartment buildings, white metal guardrails, modest palm-like greenery, an ordinary white police vehicle, several quiet onlookers, and a cordoned forensic zone farther ahead. The crime happened recently and police are still photographing evidence, organizing witnesses, and controlling access. Natural overcast daylight is the only key light. The mood advances from routine arrival to alert observation, partner reunion, entering the controlled scene, noticing evidence on the ground, and finally looking upward at an unexplained disturbance in the apartment building; never reveal the mystery itself. CHARACTERS AND REFERENCES: Character A (<<@img_8>>) is an adult Japanese female investigator with shoulder-length straight black hair, realistic skin, a dark tailored suit, white knit inner top, and a red police armband on her left upper arm. Character B (<<@img_6>>) is a middle-aged Japanese male investigator with slightly graying short black hair, tired under-eye skin, a weathered dark suit, wrinkled light shirt, and black tie. B wears NO armband. The police vehicle follows reference <<@img_7>>. Preserve each person's own facial identity, age, styling, body shape, costume and arm placement through pans, occlusions, motion blur and refocusing. Do not beautify B into a young fashion model or duplicate either lead among the extras. LENS AND EXPOSURE: Simulate a single full-frame equivalent 26mm prime lens throughout. Never zoom or change lenses. Shoot 30fps with approximately 1/60s shutter for natural motion blur. Work around T4 aperture, adjusting gently toward T3.2 only if the building shade requires it. Approximately ISO 800, fixed 5600K white balance, slightly restrained exposure protecting cloudy skies and white police car highlights, yet maintaining dark-suit texture. Camera altitude usually 1.4‒1.6m, lowered gradually to about 1.2m while passing the cordon. Move as a professionally operated human-height floating follow camera: progressive acceleration, inertia, gentle deceleration, mild handheld footfall movement, breathing and imperfect human reframing. No weightless or mechanically flawless motion. FOCUS AND LOOK: Authentic wireless follow-focus behavior shifts between A's eyes, B's face, the hand lifting the caution tape, and the evidence on the ground. Brief natural focus softness during occlusions and fast pans is acceptable, then reacquire without repeated purposeless hunting. Slight optical softness and motion blur, medium-to-deep readable depth of field. Retain visible pores, age lines, uneven real skin, natural specular oil, soft under-eye shadows, and B's subtle stubble. Restrained low-saturation cool gray, gray-green, blue-gray, charcoal and naturally pale skin; the only notable accents are the red armband, muted red cones, and yellow/black caution tape. No HDR, neon, orange/teal grading, beauty filters, artificial rim light or glowing ad-style treatment. BACKGROUND DIRECTION: Uniformed officers take separate unsynchronized actions―watch the perimeter, talk quietly, use radios. Forensic technicians independently kneel to photograph the site, position evidence markers, exchange bags and carry a case. Journalists and residents act differently; one reporter holds takeaway coffee, another checks notes, residents glance from behind the rail or lower phones. Some background people remain still. None stare at the camera, copy the leads' faces, or move in synchronized patterns. PERFORMANCE RULES: A and B naturally blink at unsynchronized intervals―roughly 3‒5 times for A and 2‒4 for B across 25 seconds, fewer during urgent attention. A's eyes lead head and shoulders; feet and weight lead directional change. Eye tension precedes changes in mouth and breath. B is about half a beat behind A's observations, eye lines and turns. A remains pragmatic and slightly tense, never performing for camera. 25-SECOND CONTINUOUS CAMERA BLOCKING: 0‒2s | Start close behind A at her right rear, about 0.5m away; her red armband enters first as she walks toward the apartments. Rise from roughly 1.3m to 1.5m and shift focus from armband to shoulder. White police car slides out on the right. 2‒4s | Follow A from one meter behind, letting upper apartment balconies and cloudy façade reveal above her shoulder while the distant caution tape becomes visible. Natural handheld micro-roll. 4‒6s | Arc smoothly around A's left side to her left-front, about one meter away. Her eyes scan the building then crowd; white railings slide across the background, onlookers move at different times, a brief shoulder occlusion leads to renewed focus on her eyes. 6‒8s | Walk backward ahead of A, leaving visual room for her gaze toward the cordon, crowd and building entrance. She slows, gathers herself, and becomes more vigilant. Reporters and distant officers continue independent activity. 8‒10s | A glances toward the right-rear, exits frame left as the camera accelerates into a motivated pan and lateral move to discover B next to the open police car door. Let the guardrail and onlookers create continuous motion blur; brake naturally and rack focus from door edge to B. 10‒12s | B lets go of the open car door, steps onto the road, straightens his worn jacket, examines the building and follows A. Camera retreats from approximately 1.5m ahead, gradually re-centering the scene to leave A room to return. 12‒14s | A walks back into frame from the left. B catches up at her right-rear by half a body length, and they approach the caution tape together. Both share a coherent frame. Technicians behind the tape photograph the ground and organize evidence bags. 14‒16s | A studies the tape height and reaches out with her RIGHT hand, back of hand first, then turning palm to lift the underside. B slows and raises his LEFT hand ready to take over. Lower camera from 1.5m to 1.4m, glance toward the hand/tape contact, then refocus on A. 16‒18s | A ducks under the tape with a physically correct right-foot-then-left-foot passage; B holds the tape with his LEFT hand and follows. The camera lowers to around 1.2m and retreats through the actual opening created by A. The yellow tape briefly blurs close to the lens. Maintain tape continuity and realistic recoil. 18‒20s | Beyond the cordon, A straightens and looks down at floor evidence; B releases the tape, which settles naturally. Camera gently arcs toward A's side/rear, tilts down and picks up evidence while A's shoulder forms a soft foreground. Technicians continue taking photos and shifting a case. 20‒22s | A's eyes rise before her chin and shoulders; she notices something above the building. B follows her attention slightly later. Camera orbits gradually to her front-right, lifts to 1.6m and racks from evidence back to A's eyes. Do not show the unseen target. 22‒24s | A stops 0.8m from lens, positioned frame right, gaze fixed upward-left; B remains roughly 1.5m behind frame left, later raising his head. Push in slightly while allowing small stabilizing motions. A's jaw firms, breath shallows. The apartment exterior occupies only the top edge of the frame. 24‒25s | Finish on A's watchful face as she holds the upward look, makes one slow natural blink, then stares. B remains perceptible behind, focused on the same unknown. Only slight handheld settling; no cut, no spoken line, no mystery reveal. SOUND: Absolutely no musical score or stylized suspense sting. Continuous grounded ambience from distant Tokyo traffic, dry leaves rustling, subtle Japanese murmurs from residents and reporters, camera shutter clicks, independently timed footsteps and clothing fabric, quiet breathing, paper rustle, takeaway-cup tapping, tape tension/recoil, forensic case sounds, and at most one or two unobtrusive distant phone notifications. Audio direction and distance must evolve smoothly with the moving camera. No sirens, police-car engine, melodramatic low-frequency hits, heartbeats, synthetic risers or dialogue from A or B. CONTINUITY / NEGATIVE CONSTRAINTS: No editing, hidden cuts, teleportation, axis reversals, changes in spatial geography, or sudden focal-length changes. A and B never swap identity, positions, faces, clothes, hairstyles, or the armband; keep A's RIGHT hand lifting the tape and B's LEFT hand taking over. Do not place the tape elsewhere, let it float, pass through bodies, snap, melt, or spontaneously move. No malformed hands, drifting feet, body intersections, duplicated faces, synchronized extras, artificial eye blinking, exposure or white-balance jumps, over-processed HDR, neon, unnecessary focus hunting, or character dialogue. This is intimate, imperfect, uninterrupted, observational Japanese crime television with a mounting sense of mystery. PROTOTYPE ASSET NOTE: The reference identifiers <<@img_8>>, <<@img_6>>, and <<@img_7>> come from the supplied production brief. The underlying reference files are not embedded in this HTML prototype.
Topview T1 garde l’image et l’histoire dans un ordre logique. Les deux clips utilisent un couloir japonais, des portes shoji et des rangées de citrouilles lumineuses. Dans T1, le fantôme sous le drap apparaît, puis les personnes réagissent une fois qu’elles peuvent le voir. Seedance 2.5 montre quelqu’un déjà effrayé avant l’apparition du fantôme, si bien que la cause de la blague arrive trop tard. Choisissez Topview T1 lorsque la scène du couloir doit rester cohérente d’un temps au suivant.
A photorealistic short set in a haunted house inside a traditional Japanese home during Halloween. Shot handheld by another staff member, with each shot about 2 seconds long. No dialogue, subtitles, or background music. Characters Sakura: a young woman playing the haunted house ghost. She is covered head to toe in a large white bedsheet with two round eye holes, the hem dragging on the floor. She stays under the sheet the whole time and never shows her head or face. The senior staffer: a man in his thirties in dark work clothes, holding a flashlight. Other guests appear only briefly in their own shots. Fixed rules There is only one ghost, Sakura. Her look never changes, and the sheet is never lifted. The closet is always on the left side of the frame, and the senior staffer hides behind a shoji door on the right. Guests always move in one direction, from the back of the frame toward the camera. The hallway is lined with jack-o'-lanterns, with warm orange light and dark surroundings. The sheet always hangs naturally and never floats. Shot order The jack-o'-lantern hallway. The closet on the left slides open and the bedsheet ghost jumps out, spreading her sheet up high with both hands. A couple glances at her; the guy even pulls out his phone for a selfie with her, and they walk past laughing. The ghost lowers her arms and retreats into the closet, disappointed. Inside the closet, the bedsheet ghost stomps in frustration, her fists clenched under the sheet. Two housewives walk past, chatting. The ghost rushes out, but steps on her hem and stumbles. By the time she steadies herself, they're already far away, and she can only raise her sheet at their backs. The senior staffer comes over, pats her on the shoulder, points to the shoji door on the right, and flashes a flashlight twice to show her the cue. Inside the closet, the ghost's shoulders rise and fall with a deep breath. She hitches up her hem and gives a determined nod. Through the gap in the closet door, two little boys wearing devil horns and capes timidly approach, holding hands. Behind the shoji door on the right, the flashlight flashes once, then twice. The ghost bursts out of the closet and spreads her sheet up high. The two boys scream, jump, and cling to each other. The boys turn and run away, and one of them drops his devil horns on the floor. Left behind, the ghost is startled by their screams and falls back onto the floor. The sheet stays over her head, and she trembles all over. The curtain on the shoji door lifts again. Several staff members peek out with wide eyes, then silently applaud her. The ghost picks up the devil horns and shyly puts them on top of her sheet. She takes the senior staffer's hand to get up, gives a happy little wiggle, and returns to the closet.