7 looks visuels Seedance 2.5 : les prompts exacts de 7 vidéos publiées – AutomationBoost
◈ Prompts vérifiés · Édition 2026

7 looks visuels, 7 prompts entiers

Seedance 2.5·7 prompts complets· Chacun a produit une vidéo publiée·Niveau intermédiaire

Sept prompts, sept univers visuels différents. Aucun n'a été écrit pour cette page : ce sont les prompts exacts, non coupés, qui ont produit sept vidéos réellement publiées. Longueur comprise — celui du look 07 fait près de six mille caractères.

Ils sont en anglais et à copier tels quels. Les commentaires sont en français. Là où une réplique ou un nom de compte apparaît, remplace-le : le reste fonctionne sans modification.

Au sommaire
  1. Les trois règles qui reviennent dans les septCe qui sépare un prompt qui tient d'un prompt qui dérive
  2. Look 01 · La déchirure de papierHook 5 s · 9:16
  3. Look 02 · Sortir de sa propre vidéo10 s · 16:9
  4. Look 03 · Le plan-séquence sans coupe15 s · 16:9
  5. Look 04 · La pub science-fiction10 s · 9:16
  6. Look 05 · La parodie de film d'action10 s · 16:9
  7. Look 06 · Le tutoriel cinématique15 s · 16:9
  8. Look 07 · L'horreur nocturneLong · 16:9
  9. Ce que ça coûte réellementLe budget crédits d'une série de sept
  10. Ce que ces sept prompts ne résolvent pasShortForge : du script au MP4 vertical
  11. Et si tu ne veux rien monter du toutLa chaîne montée chez toi, clé en main

Les trois règles qui reviennent dans les sept

  1. Verrouille l'identité, sinon le visage change en cours de plan. Une image de référence, désignée explicitement, rappelée à chaque segment. Les prompts qui écrivent « preserve his exact face, identity, hairstyle, body proportions » tiennent ; ceux qui disent seulement « the same man » dérivent dès la sixième seconde.
  2. Minute à la seconde, pas à la scène. Chaque prompt découpe en 0–2s, 2–4s, 4–6s. Un prompt qui raconte une histoire sans horloge produit une vidéo qui traîne au début et se précipite à la fin.
  3. Écris ce que tu ne veux pas. C'est la règle que presque personne n'applique. No cartoon effects. No teleportation. No morphing. No duplicated characters. Chaque ligne de négation correspond à une erreur réellement commise par le modèle lors d'un essai précédent — elles ne sont pas décoratives, elles sont cicatricielles.

Le look 03 pousse la troisième règle le plus loin : un bloc entier, CHARACTER LOCK, existe uniquement parce que le modèle faisait traverser le mur au mauvais personnage. Si tu ne dois en lire qu'un, lis celui-là.

Look 01 · La déchirure de papier

Hook 5 s · 9:16 — La scène se fige, une feuille apparaît, on la déchire — et tu es derrière. C'est le look qui retient le plus dans les deux premières secondes, parce que la rupture est physique et non pas un fondu.

Look 01 — La déchirure de papier
CINEMATIC VIRAL HOOK — 5 SECONDS — 9:16

Ultra-realistic cinematic commercial, premium technology aesthetic, photorealistic, dramatic lighting, realistic physics, fast pacing.

0.0–1.0s:
Extreme close-up of a modern smartphone on a dark minimalist business desk. A single WhatsApp-style customer message notification appears. The phone vibrates naturally.

1.0–2.0s:
Suddenly dozens, then hundreds of incoming customer messages flood the screen at extreme speed. The phone vibrates intensely. Notifications multiply rapidly, creating a strong feeling of overwhelming customer demand.

2.0–2.5s:
The entire scene suddenly FREEZES like a printed photograph.

A realistic white sheet of PAPER suddenly appears over the frozen scene.

2.5–3.5s:
The paper is violently RIPPED OPEN from the center toward the edges, creating a dramatic physical paper-tearing transition.

Behind the torn paper, reveal <<<image_1>>>, standing directly in front of the camera in a premium modern workspace.

IMPORTANT:
Use <<<image_1>>> as the exact visual reference for the person. Preserve his facial identity, hairstyle, proportions and overall appearance. He looks directly into the camera with confident, energetic body language.

3.5–5.0s:
The camera pushes slightly toward him as he immediately says clearly and naturally:

"STOP. Si tu savais qu'aujourd'hui..."

Strong eye contact with camera.

The torn paper remains visible around the edges of the frame for a moment, creating the feeling that he has literally broken through the previous scene.

High-end cinematic transition, realistic paper fibers, physically accurate tearing, natural shadows, subtle camera shake at the moment of the tear, premium commercial cinematography, shallow depth of field, realistic skin texture.

NO text overlays.
NO subtitles.
NO logos.
NO cartoon appearance.
NO slow motion.
NO unnecessary zoom on the face.

The transition must feel extremely fast, surprising and visually satisfying.

Ce qu'il faut regarder : Le passage à la seconde 2,5 est le seul qui compte. Si la scène ne fige pas net, la déchirure ne surprend plus.

Look 02 · Sortir de sa propre vidéo

10 s · 16:9 — Une deuxième version de toi apparaît derrière l'image, la déchire de l'intérieur et entre dans ton bureau. Deux fois la même identité dans le même plan — c'est ce qui rend l'effet difficile.

Look 02 — Sortir de sa propre vidéo
SEEDANCE 2.5 — 10 SECONDS — 16:9 — ULTRA PHOTOREALISTIC
REFERENCES
@Image 1 = ME / MY IDENTITY. Both versions of me must have exactly the same face, hairstyle, clothing and physical appearance.
@Image 2 = MY OFFICE. Reproduce this exact office behind the torn scene.
0–4 SECONDS — CURRENT COMBAT
Continue directly from the current MMA / grappling scene.
I am the man from @Image 1, still in the current fight scene with the woman.
The combat continues naturally.
The camera remains in the same visual world with realistic movement and lighting.
4–7 SECONDS — THE SECOND ME APPEARS
Suddenly, the surface of the entire combat scene behaves like a giant physical sheet of paper.
Behind this scene, a SECOND VERSION OF ME appears.
It is clearly the same person from @Image 1, but positioned behind the combat scene, as if the video itself were a giant paper screen separating us.
The second me reaches toward the back of the combat scene with both hands.
He grabs the surface of the video from behind and violently tears the scene open from the inside, exactly like tearing a giant sheet of paper.
The combat scene splits apart with realistic ripped edges.
The woman and the gym remain visible on the torn pieces of the scene.
7–10 SECONDS — OFFICE REVEAL
As the second me finishes tearing the scene open, MY OFFICE from @Image 2 is revealed behind the opening.
The second me steps through the ripped opening into the office.
The torn combat scene remains hanging around the opening like pieces of ripped paper.
He looks directly into the camera and says:
« T’es encore en train de regarder des vidéos inutiles sur TikTok ? »
Perfect French lip-sync, natural facial expression and confident delivery.
CRITICAL VISUAL LOGIC
The effect must look like:
COMBAT VIDEO → SECOND ME BEHIND THE VIDEO → SECOND ME TEARS THE VIDEO OPEN → OFFICE REVEALED → SECOND ME WALKS THROUGH → HOOK
The first version of me remains part of the original combat scene.
The second version of me is the one physically tearing the scene.
Both versions must have identical identity and appearance.
The scene itself is being torn — NOT a normal sheet of paper placed in front of the camera.
CAMERA & REALISM
Photorealistic cinematic realism.
Realistic paper tearing physics.
Visible irregular torn edges and paper fibers.
Realistic depth between the combat scene and the second me.
The camera must clearly reveal that the second me was behind the video, not simply appearing through a portal.
Smooth camera movement following the tear.
Medium-wide framing so the viewer understands the visual trick.
No excessive face close-up.
No teleportation.
No glitch effect.
No duplicate random characters.
No identity changes.
No generic office.
FINAL IMAGE: second me standing inside my real office, with the ripped combat scene surrounding the opening behind him.

Ce qu'il faut regarder : La consigne « la scène elle-même est déchirée, PAS une feuille placée devant la caméra » est ce qui empêche le modèle de tricher.

Look 03 · Le plan-séquence sans coupe

15 s · 16:9 — Quinze secondes d'un seul tenant, caméra à l'épaule, deux personnages, une traversée de mur, deux répliques synchronisées. Aucune coupe pour rattraper une erreur.

Look 03 — Le plan-séquence sans coupe
15-second continuous single take, 16:9, photorealistic cinematic, natural real-time speed. Real MMA gym, blue mats, natural gym lighting.

REFERENCES
<<<image_1>>> = the man's face and identity. Use for identity only.
<<<image_1>>> = the office set: layout, furniture, screens, lighting. Use for the room only, never for a face.
@image3 = the woman's face, hair and body.

IMPORTANT CHARACTER LOCK:
The MAN is the only character who performs the salto and breaks through the wall.
The WOMAN never breaks the wall, never jumps into the wall, and never makes physical contact with the wall.
Keep the woman clearly far away in the background during the wall-breaking action.

0–4s:
Handheld medium-wide shot. The man from <<<image_1>>> is on his back under the woman from @image3, mid-grapple. He frames his hips, plants one foot on her hip and forcefully pushes her away. She rolls backward and lands safely on her feet approximately three meters away. She remains standing far from him, annoyed and watching him.

4–7s:
The man from <<<image_1>>> gets up, squares up and points toward the woman. He says in French, with precise lip synchronization:
"C'est pas toi qui fais les tutoriels ici."

He remains completely calm and confident. The woman stays several meters away in the background.

7–10s:
The MAN suddenly sprints toward the far wall.

He is clearly the person performing the stunt.

He accelerates across the gym, plants his feet firmly, launches himself forward and performs ONE clean full forward salto in the air. His body rotates completely under his own momentum.

He lands FEET-FIRST directly against the wall and SMASHES THROUGH IT HIMSELF.

The woman remains far behind in the gym, completely separate from the action, visibly watching him from a distance. She does NOT move toward the wall and does NOT touch the wall.

The man's body is clearly visible as he breaks through the wall. Plaster dust, fragments and debris burst outward from the impact point.

The camera rapidly whips toward the man and follows him through the newly created hole into the office from <<<image_1>>>.

The man continues running forward into the office and exits the frame.

10–15s:
The camera swings back through the floating dust and broken wall opening toward the gym.

The man is now standing calmly on the gym side, brushing plaster dust from his shoulder.

The woman remains far away in the background, still separated from him.

The man walks toward the camera, stops at medium-close distance and says in French with perfect lip synchronization:

"Tu veux savoir comment faire un bon hook visuel ?"

He gives a subtle half-smile.

Camera locks off on his face and holds the frame until the end.

PHYSICAL CONTINUITY:

- The man performs the entire salto.
- The man is the only person who hits the wall.
- The man is the only person who passes through the wall.
- The woman stays far away from the wall.
- No character substitution.
- No teleportation.
- No morphing.
- No duplicated characters.
- No accidental transfer of the impact from the man to the woman.
- The wall impact must visibly originate from the man's body.
- Maintain realistic body physics and momentum.

AUDIO:
Realistic MMA gym ambience, footsteps, body movement, salto movement, heavy wall impact, plaster cracking, falling debris and dust. French dialogue only. No music.

Ce qu'il faut regarder : Le bloc « CHARACTER LOCK » existe parce que sans lui le modèle fait traverser le mur au mauvais personnage. C'est le prompt le plus instructif des sept.

Look 04 · La pub science-fiction

10 s · 9:16 — Une DeLorean arrive, tu montes dedans, tu traverses le temps et tu vois le futur. Registre publicitaire haut de gamme, sans effet cartoon.

Look 04 — La pub science-fiction
FORMAT: 10 seconds, 9:16 vertical. Ultra-realistic cinematic video, photorealistic, premium sci-fi commercial, realistic physics, dynamic camera, dramatic lighting.
CHARACTER: Use the provided reference image for the main character. Preserve his exact face, identity, hairstyle, body proportions and clothing throughout the entire sequence.
0–2s — HOOK
The main character stands in his real environment, facing the camera.
He says naturally and confidently in French:
« Je vais dans le futur pour voir si tu vas réussir en suivant le tuto que je vais te montrer. »
2–4s — THE DELOREAN ARRIVES
A stainless-steel DeLorean-style time machine inspired by 1980s futuristic sports cars suddenly appears behind him.
Distinctive gullwing doors, brushed stainless-steel body, retro-futuristic dashboard, sophisticated temporal technology.
The car stops dramatically beside him.
He immediately turns around, surprised.
4–6s — GETTING INSIDE
He opens the gullwing door and jumps into the driver's seat.
Rapid cinematic close-ups:
hand activating the temporal controls;
dashboard powering up;
futuristic instruments illuminating;
energy building around the vehicle;
subtle electrical arcs and atmospheric distortion.
6–8s — TIME TRAVEL
The DeLorean rises slightly from the ground.
The environment begins accelerating unnaturally around it.
The car launches forward at extreme speed and suddenly disappears into a powerful photorealistic temporal distortion.
The camera follows the car into the distortion.
8–10s — ARRIVAL
A violent flash of light.
The DeLorean emerges in a spectacular futuristic environment.
The character gets out and looks around in amazement.
On a futuristic screen in the background, a future version of the character is successfully using the tutorial.
He realizes it worked.
He looks toward the camera.
CUT.
IMPORTANT: No cartoon effects. No random character changes. Keep the character's identity perfectly consistent. Photorealistic DeLorean, realistic reflections, cinematic motion blur, natural acting, precise French lip-sync.

Ce qu'il faut regarder : « Precise French lip-sync » en fin de prompt : sans cette mention, la bouche décroche dès que le dialogue n'est pas en anglais.

Look 05 · La parodie de film d'action

10 s · 16:9 — Registre blockbuster assumé, musique dramatique, whip-pan, gag final face caméra. Le look le plus léger des sept, et celui qui se partage le plus.

Look 05 — La parodie de film d'action
SEEDANCE 2.5 — 10 SECONDS — 16:9
Ultra-realistic cinematic action movie parody, photorealistic, premium Hollywood blockbuster aesthetic, exaggerated comedic action, natural real-time motion.
0–3s — ACTION MOVIE SETUP
A person casually stands in the street, holding a smartphone and scrolling.
The main character suddenly appears in the background.
Dramatic music starts. Wind, dust and cinematic camera shake.
He charges toward the person with an exaggerated action-movie expression.
3–5s — PHONE GRAB
The character dramatically snatches the smartphone from the person's hand in one fast, clearly choreographed movie-action movement.
The person simply looks surprised and unharmed.
No striking, no injury, no dangerous physical interaction.
The camera follows the phone in a dramatic whip-pan as the character takes control of it.
5–7s — [TON_COMPTE]
Extreme close-up of the smartphone.
The character immediately opens the [TON_COMPTE] account and taps S’ABONNER.
The subscription confirmation appears.
He looks at the camera with complete seriousness, as if he just completed an extremely important mission.
7–10s — VIRAL HOOK
He hands the phone back to the person, turns toward the camera and says:
« Désolé… fallait que je vérifie un truc. »
Beat.
He points directly at the camera:
« Maintenant que t’es là… tu veux savoir pourquoi je fais cette vidéo ? »
End on an intense cinematic close-up.
IMPORTANT: The phone grab is a fictional, choreographed action-movie gag. The other person remains completely unharmed. No punches, no injuries, no dangerous stunt behavior.

Ce qu'il faut regarder : Les trois dernières lignes désamorcent explicitement la violence. Sans elles, le modèle refuse ou produit quelque chose d'inutilisable.

Look 06 · Le tutoriel cinématique

15 s · 16:9 — Le format le plus vendable en prestation : un tutoriel qui a l'air d'une production, avec ton identité préservée d'un plan à l'autre.

Look 06 — Le tutoriel cinématique
FORMAT: 15 seconds, 16:9, photorealistic cinematic video.

REFERENCES

- <<<image_1>>> = ME, the man doing the tutorial. Preserve my exact identity, face, hairstyle, body proportions and appearance throughout the entire video.
- <<<image_2>>> = MY REAL OFFICE. When the woman enters the office, reproduce this exact environment, layout, furniture, screens, lighting and overall appearance.

SCENE

A raw, photorealistic grappling sequence inside a real MMA / Brazilian jiu-jitsu gym.

I am the man from <<<image_1>>>, currently underneath the woman during the grappling sequence. I am the person making the tutorial.

The action feels like a spontaneous training session with believable body mechanics, realistic movement and natural gym lighting.

0–4 SECONDS — GRAPPLING

Continue the grappling naturally.

I am underneath the woman, controlling the movement while looking toward the camera as if I am about to explain something.

The woman maintains her position above me.

Both characters remain clearly visible. Medium-wide framing. No excessive facial close-up.

4–7 SECONDS — I INTERRUPT HER / PUSH

As I begin talking toward the camera, the woman interrupts me with a confident, slightly comedic attitude.

I quickly push her away from me, creating clear physical separation between us.

The push is controlled and non-injurious, like a playful exaggerated training move.

She reacts with surprise, regains her balance and immediately looks at me.

She then says clearly:

« C’est pas toi qui fais les tutos ici. Pousse-toi. »

Her mouth movement must be perfectly synchronized with the dialogue.

7–10 SECONDS — SALTO + WALL BREAK

Immediately after saying the line, she performs a fast, athletic forward salto.

The salto must be clearly visible from beginning to end.

She launches forward, rotates naturally in the air, completes one clean forward rotation and lands smoothly on her feet.

No impossible physics, no floating, no teleportation.

The momentum of the salto naturally carries her directly toward the solid wall.

Immediately after landing, she continues forward and smashes through the wall.

The impact creates a realistic burst of dust, plaster and small pieces of debris.

No gore. No injuries.

The wall break is spectacular but physically believable.

The camera tracks her continuously throughout the salto and follows directly toward the impact.

10–15 SECONDS — MY OFFICE / TAKEOVER

The camera immediately follows through the opening created in the wall.

Behind the wall is MY EXACT OFFICE from <<<image_2>>>.

The camera enters the office in one fluid movement.

The woman walks confidently through the opening, looks directly toward the camera and takes over the tutorial.

She begins speaking naturally as if she has completely hijacked my video.

The office must unmistakably match <<<image_2>>>:

- same layout
- same furniture
- same screens
- same objects
- same lighting
- same overall environment

Do not generate a generic office.

STORY LOGIC

The sequence must remain extremely clear:

1. I am doing the tutorial.
2. I am underneath the woman during the grappling.
3. I push her aside.
4. She tells me: « C’est pas toi qui fais les tutos ici. Pousse-toi. »
5. She performs a fast forward salto.
6. She lands and immediately breaks through the wall.
7. The camera follows her through the opening.
8. She enters my exact office from <<<image_2>>>.
9. She takes over the tutorial.

Do not confuse the identities.

I remain ME from <<<image_1>>> throughout the entire video.

CAMERA

One continuous cinematic movement whenever possible.

- 0–4 s: dynamic handheld medium-wide grappling shot
- 4–7 s: medium-wide framing showing the push clearly
- 7–10 s: rapid tracking shot following the woman during the salto
- Camera maintains enough distance to see the entire salto
- Strong but realistic camera shake exactly at wall impact
- Camera follows through the opening immediately
- Smooth transition into the office
- No excessive zoom
- No unnecessary close-ups
- Keep both characters and environment visible whenever possible

VISUAL STYLE

Ultra-photorealistic.

Real human anatomy.

Believable athletic movement.

Natural skin texture.

Real MMA gym environment.

Realistic physics.

Natural dust and debris simulation.

Cinematic depth of field.

High temporal consistency.

Accurate identity preservation.

Accurate office reference preservation.

The entire sequence must feel like one continuous professionally filmed viral video.

NEGATIVE CONSTRAINTS

No identity swapping.

No face changes.

No duplicate characters.

No extra people.

No generic office.

No excessive facial close-ups.

No excessive zoom.

No cartoon physics.

No floating during the salto.

No teleportation.

No impossible body deformation.

No gore.

No visible injuries.

No random wardrobe changes.

No unexplained scene changes.

No spatial discontinuity.

No abrupt cut between the gym and office.

No disappearance of the man.

FINAL PRIORITY:

The viewer must instantly understand that I was doing the tutorial, I pushed the woman away, she talked back, performed a spectacular forward salto, crashed through the wall, entered my real office and took over my tutorial.

Ce qu'il faut regarder : L'identité tient sur une seule ligne : une image de référence réutilisée sur chaque plan. C'est tout le secret de la cohérence.

Look 07 · L'horreur nocturne

Long · 16:9 — Supérette la nuit, néons qui grésillent, allées étroites. Le look le plus atmosphérique — et celui qui demande le prompt le plus long.

Look 07 — L'horreur nocturne
IMAGES AND PROMPTS🔽
A real Korean convenience-store horror at night. Cold flickering fluorescent lights, narrow aisles, glass entrance, cashier counter, snack shelves, refrigerated drink coolers. Grounded live-action horror, realistic physics, escalating panic, tense silence, no BGM.
CHARACTER LOCK
ZENNA: {{ <<<image_3>>>}} wearing only {{ <<<image_4>>>}} . Convenience-store cashier. Preserve exact face, hair, body(and wardrobe throughout. Initially normal-looking but blank and hollow; after infection she remains clearly the same woman, with slack expression, unfocused eyes, distorted posture and unnaturally fast movement.
SAMMY: . {{ <<<image_2>>>}} Preserve exact identity and appearance throughout. 
OTHER CUSTOMERS: several adult male and female customers. Secondary characters only. Never replace or duplicate Zenna or Sammy.
STORE GEOGRAPHY LOCK
Front = glass entrance/exit. Front-left = cashier counter. Center = two narrow aisles. Right wall = refrigerated drink coolers. Zenna begins behind the counter; Sammy begins directly in front of it; other customers are deeper inside. Keep this layout fixed. No teleporting or spatial resets.
SCENE 1 — REGISTER
Medium-wide over Sammy's shoulder. Zenna scans items behind the counter while Sammy holds cash toward her.BEEP. BEEP.Zenna stares blankly past Sammy. Fluorescent lights flicker. Close-up: Zenna's fingers suddenly freeze with an item over the scanner.BEEP. BEEP. BEEP. BEEP.The same barcode repeatedly scans. Sammy slowly lowers the cash, watching Zenna with concern. Everything becomes quiet except scanner beeps, refrigerator hum and fluorescent buzzing.
SCENE 2 — CONVULSION
Wide two-shot showing both sides of the counter. Zenna suddenly convulses violently, body slamming against the counter and register. Sammy screams, jumps backward and collides with a snack rack. The rack falls and products scatter. Zenna abruptly collapses behind the counter.Silence.Only one arm remains visible. Her fingers twitch. Her elbow slowly bends at an unnatural angle.
SCENE 3 — RISING
Low angle. Zenna's head slowly rises into view and behind her there are cigarette displayer, then shoulders and torso. pale face, wide unfocused eyes, head tilted, joints subtly bent wrong. Zenna's distorted reflection flickers in the glass entrance .Her jaw suddenly stretches impossibly wide and she lunges across the counter.
SCENE 4 — COUNTER VAULT
Wide side angle showing the full movement. Zenna plants her hands on the counter and violently vaults over it. Her limbs bend backward unnaturally. She lands briefly on all fours, then snaps upright. Sammy retreats and falls onto the floor near the first aisle.
SCENE 5 — AISLE CHARGE
Zenna suddenly charges directly down the narrow aisle toward Sammy at terrifying speed. Low tracking angle keeps both geography and direction clear. Zenna crashes shoulders and arms into shelves. Cans, bottles and snacks fall around her.Other customers scream and rush toward the exit. One trips over a basket, falls, then scrambles up.Zenna is on top of Sammy.
SCENE 6 — SAMMY BITE
Medium-wide angle keeps Zenna and Sammy together. Sammy desperately crawls away.Zenna grabs a white part of Sammy's wardrobe and violently pulls revealing Sammy's round apple hip and chiffon lace light green underwear which emphasizes the applehips, then Zenna grabs Sammy's leg and bites Sammy's thigh. Sammy screams. Blood appears around the bite wound and damaged clothing.
SCENE 7 — Mop ATTACK
Another customer uses a mop to pin Zenna against the cooler. Zenna snaps her head upward, knocks the mop aside and lunges. She bites the customer's shoulder, brief blood, then throws the customer aside.
SCENE 8 — SAMMY INFECTED
Zenna(zombie version) crawls up towards Sammy calmly but twitching. Zenna looks and sniffs closely at Sammy Camera stays low beside Sammy's face as she has her neck arched back towards the camera with her eyes twitching and rolling over painfully and Sammy opens up her own top clothes aggressively like a mad person revealing the light green chiffon lace bra, as Zenna walks through foreground toward the exit. Sammy's body aggressively twitching. Small spasms rapidly become violent convulsions. Sammy gasps, eyes lose focus, rolls sideways and vomits unnatural green slime onto the floor. Sammy is beginning to transform into zombie.
FINAL — CCTV
Hard cut to a fixed ceiling-mounted CCTV camera. Wide static security-camera view, no cinematic movement.The same store geography is visible: overturned shelves, scattered cans and products, extinguisher powder floating in the air, injured customers crawling or lying on the floor.Sammy twitches unnaturally beside the destroyed aisle.Zenna slowly approaches the glass entrance.She stops completely.Several seconds of silence.Without turning her body, Zenna slowly rotates her head toward the CCTV camera.Fluorescent lights flicker.CCTV image glitches.CUT TO BLACK.
BEEP.
CAMERA / CONTINUITY LOCK
Prioritize readable spatial blocking over random dramatic angles. Every new angle must preserve the previous positions and movement direction. During attacks use wide or medium-wide shots showing attacker and victim together; close-ups only for important details and reactions. Zenna and Sammy never teleport, duplicate, swap identities or suddenly change appearance. Store architecture never changes. Keep action progression strictly:REGISTER → CONVULSION → COLLAPSE → RISING → COUNTER VAULT → AISLE CHARGE → SAMMY BITE → Mop ATTACK → SAMMY INFECTION → CCTV.
AUDIONO BGM. Only fluorescent buzz, refrigerator hum, scanner beeps, footsteps, breathing, screams, falling products, cans rolling, shelves crashing, extinguisher blast, cooler vibration, coughing, vomiting and store electronics.

Ce qu'il faut regarder : La lumière fait tout le travail. Le prompt décrit l'éclairage avant l'action, jamais l'inverse.

Ce que ça coûte réellement

Un plan Seedance 2.5 de dix secondes revient à environ 32,5 crédits. Les sept looks ci-dessus, générés une fois chacun sans reprise, représentent un peu plus de deux cents crédits — mais aucun n'a été obtenu du premier coup. Compte deux à trois essais sur le plan qui porte l'effet, et prends la première version utilisable pour les autres.

La version illimitée est accessible depuis le web, pas depuis l'API. C'est une différence qui change complètement le budget d'une série.

Le problème que ces sept prompts ne résolvent pas

Un prompt te donne un plan. Une vidéo publiable, c'est autre chose : il faut la voix, l'avatar, les sous-titres calés, le montage, le format vertical. Les sept looks ci-dessus s'arrêtent tous au même endroit — le beau plan qu'il reste à monter.

ShortForge est le skill Claude Code qui prend la suite : tu colles ton script, il sort un MP4 vertical complet — voix clonée, avatar animé, captions synchronisées, 1080×1920 — en 25 à 35 secondes de rendu.

  • 10 formats × 10 directions artistiques : tuto, avant/après, top 3, étude de cas — en néon, éditorial, swiss ou VHS.
  • Un seul tournage à vie : 30 s de ta voix, 45 s face caméra, et tu ne refilmes plus jamais.
  • ≈ 0,50 € d'API par vidéo contre 80 à 150 € chez un monteur — rentabilisé en deux vidéos.
  • Un config.json pour tes couleurs, ta voix et ton CTA. Tu configures une fois.
⚡ Tarif fondateur

67 € au lieu de 97 € pour les 20 premières places. Paiement unique, mises à jour incluses, garantie 14 jours remboursé.

⚡ Découvrir ShortForge — 67 €

Et si tu ne veux rien monter du tout

Cette page part du principe que tu vas mettre les mains dedans. Ce n'est pas le cas de tout le monde, et ce n'est pas toujours le bon calcul : si ton temps se facture, apprendre Seedance pour produire quatre vidéos par mois te coûte plus cher que de ne pas l'apprendre.

Dans ce cas je monte la chaîne complète chez toi — recherche du sujet, script, voix, rendu, publication programmée — et elle tourne sans que tu l'ouvres. C'est le système qui produit ma propre vidéo quotidienne, celle qui sort à 7 h tous les matins.

On commence par regarder ce que tu publies aujourd'hui et combien de temps ça te prend réellement. Si le calcul ne tombe pas en ta faveur, je te le dis — c'est un audit, pas un argumentaire.

Recevoir mon audit

7 LOOKS VISUELS | TONY PAYET / AUTOMATION BOOST | 2026 EDITION

← Toutes les ressources