Seedance 2.5拍完整电影的视频提示词
27段分镜+韩语唇同步，30秒犯罪短片一键出片
今天两条提示词都在榨干 Seedance 2.5 的叙事上限。一条是 30 秒韩式犯罪惊悚，从分镜时间码到唇同步规则写满 11 场戏；另一条走伪纪录片路线，用廉价 DV 质感拍森林 cryptid 目击，呼吸声、失焦、过曝全算进提示词里。两条路子截然不同，核心共同点：把"视频模型当剧组用"，提示词本身就是制片手册。
@doctorwasif 的提示词(⭐ 27 人收藏)
11 场戏带时间码、韩语对白规则、道具连续性检查——这不是 prompt，是直接用英文写的分镜脚本，存了改个场景就能复拍。
Now i can make a complete film with Seedance 2.5 on @openart_ai
Prompt:
Use <<image>>only for the WOMAN’s face, identity, hair, skin texture, and body proportions. Do not copy its background, clothing, lighting, or composition.
30s cinematic thriller, photorealistic. Rainy gas-station mini-mart after a car crash. Controlled handheld camera, fluorescent practicals, intermittent red/blue police lights, wet asphalt, haze, realistic sweat/grime, shallow DOF, film grain. No music.
CAST: WOMAN throughout: torn dark jacket, jeans, scraped forearm, messy half-loose hair, sweaty, aggressive and adrenaline-fueled. CLERK: tired 40s man, green vest/flannel, cautious. STRANGER appears only in S9, dark silhouette, face never visible. No other people.
DIALOGUE RULE: Korean, natural quiet speech, perfect lip-sync. Action must completely stop before speaking; speaker faces camera/front 3/4 with unobstructed mouth and remains still during the line. No VO/off-screen speech. S9 silent.
S1 0–2.5: Woman bursts in, grabs unlabeled water from cooler, rushes to counter.
S2 2.5–5: Slams bottle down, stops, faces camera: “붕대 있어요? 빨리요.”
S3 5–7.5: Clerk cautiously gets bandage kit, stops, faces camera: “저기… 괜찮으신 거예요?” Places kit down.
S4 7.5–9.5: Woman grabs kit, stops: “신경 꺼요.” Clerk backs away.
S5 9.5–12: She sits by window, opens kit and wraps forearm. Police lights/wreck briefly visible outside. No dialogue.
S6 12–14.5: Finishes bandage, stops, faces camera: “…살아있네.” Then drinks water.
S7 14.5–16.5: Watches police lights through rainy glass, sets bottle down. No dialogue.
S8 16.5–18.5: Hears something, clenches fist, rises and faces door. No dialogue.
S9 18.5–22: Stranger bursts in and grabs her. Brief raw struggle—she twists free, elbows/shoves him, shelf rattles, bottle rolls, she forces him outside. His face never visible. No dialogue.
S10 22–25: She reaches door, stops, looks back at clerk: “문 잠가요. 아무도 안 들여보내고.” Then exits.
S11 25–30: She sprints through rain, jumps into dark sedan, starts instantly, reverses and speeds away from the police and crash. Police car passes the now-empty station too late. Door remains open/swinging.
CONTINUITY: Same woman/face, store layout, door, red OPEN sign, water bottle and bandage progression throughout. No duplicated props/people, no weapons/blood/gore, no subtitles/text/logos/watermarks, no beauty retouching, no slow motion/speed ramps.
@CharaspowerAI 的提示词(⭐ 17 人收藏)
DV 粗粝感不是后期加的，是提示词里一条条"坏画质指令"堆出来的：压缩噪点、自动对焦失灵、麦克风贴着呼吸——存了改个法语台词就能换国家拍。
POV: your cheap camcorder records something that shouldn't exist... 👀🌲
This is one of the most convincing found footage clips I've created with Seedance 2.5.
PROMPT👇
Ultra-realistic amateur cryptid sighting footage. Cheap handheld camcorder look, low-quality image, unstable zoom, bad autofocus, heavy digital grain, compression artifacts, shaky framing, harsh breathing close to the microphone, accidental overexposure through the trees, raw found-footage realism. No polished cinema look.
One unseen camera operator only, frightened adult hiker breathing heavily behind the camera. One massive Bigfoot-like forest creature: tall, broad, covered in dark matted hair, humanlike posture, heavy natural movement, never cartoonish.
Dense forest in daylight, uneven ground, wet leaves, tree trunks, foggy depth, broken branches, deep natural shadows.
Authentic low-grade camcorder texture, believable forest sound, realistic creature anatomy, limited but clear reveal, unstable autofocus, strong viral payoff, natural movement, terrifying realism, no comedy, no clean digital polish.
0-5s: Shaky handheld POV - The camera moves through the forest, breathing loud, branches hitting the lens. The operator whispers in French: “Attends… attends… c’est quoi ça là-bas ?” The image struggles to focus between trees.
5-10s: Messy zoom-in - Through the trees, a huge hairy figure is barely visible in the distance, half-hidden behind a trunk. The operator mutters: “Non… non… c’est pas possible…” The creature slowly steps forward into a small patch of light.
10-15s: Unstable push and panic reframing - The Bigfoot-like creature turns its head toward the camera. Clear view for only a second: massive shoulders, long arms, wet dark fur, heavy breathing. The operator gasps and whispers: “Oh mon Dieu…”
15-20s: Violent shake, partial loss of frame - The creature suddenly lets out a deep roar and takes two fast steps toward camera. The operator panics, stumbles backward, drops the camera toward the ground. The last image catches the creature rushing closer through the trees before the footage cuts to black.
#DreaminaCPP @dreamina_ai