Seedance 2.5 提示词
分享链接 · https://openimages.ajiang.me/p/2108050694146335188/

composition · centeredcomposition · close-uphuman_attributes · headwearhuman_attributes · long-hairhuman_attributes · makeuplighting · soft-light
提示词(原文,逐字保留)
Seedance 2.5 Prompt
Create a NEW 15-second Japanese action-comedy with synchronized dialogue and sound effects. Use the selected aspect ratio.
STORY
A buys ONE soda. The vending machine suddenly pours out hundreds of cans. A and B flee, but the rolling avalanche catches and buries them up to their shoulders. B calmly opens one can. End on A's familiar stunned face, holding the single can she bought.
REFERENCES
A = the black-haired character from @Image 1; voice timbre from @Audio 1.
B = the white-haired character from @Image 2; voice timbre from @Audio 2.
Preserve exact faces, hairstyles, costumes, accessories, proportions and reference art styles. References define identity and voice, not poses, backgrounds or scripts. Never merge identities or replay recorded words.
SCENE AND STYLE
Deserted sidewalk beside Shibuya Scramble Crossing at dusk. One human-scale vending machine stands screen-right, its lower retrieval opening facing camera. A stands beside it; B waits to her left. Payment is already complete.
Photorealistic Shibuya architecture, crossing stripes, pavement, machine, cans and lighting. A and B remain hand-drawn 2D anime with clean lines and cel shading, integrated with contact shadows and reflected light.
Rows of plain drink cans and selection buttons; no readable labels or logos.
PROP AND ACTION CONTINUITY
All cans are hand-sized metal cylinders, silver with a cyan band and pull tab. No liquid flood, foam, giant cans or shapeless metallic mass.
Every new can emerges from the SAME retrieval opening. The machine keeps its size and design.
Cans shoot out in a narrow stream, fan outward, bounce and roll into a ground-level avalanche that catches the pair.
A receives the FIRST can in her RIGHT hand and keeps that same unopened can through the final frame. She never presses the button a second time.
Only TWO characters; BOTH are caught. Keep their faces visible and bodies uninjured. Retain the settled pile around them between shots.
0.0-2.0s — JUST ONE DRINK
Open on a medium two-shot showing both characters, the selection button and the retrieval opening.
A presses ONE button with her left hand. At 0.2-1.1s, A says in Japanese {いっぽんだけ。}, voice timbre referencing @Audio 1, with natural synchronized lips.
One can drops into the tray with an ordinary clunk. A picks it up with her right hand. B watches casually.
Finish the purchase by 2.0s; no walk-up or coin insertion.
(Light, brief instrumental everyday music.)
<Button click, one ordinary can clunk.>
2.0-4.5s — THE MACHINE WILL NOT STOP
Hold a clear three-quarter view including the machine, both characters and the ground.
Immediately after A takes her can, another drops, then several more. The clunks accelerate into a rapid mechanical rattle.
By approximately 2.8s, a continuous stream of cans bursts from the retrieval opening. It fans outward and becomes a rolling avalanche across the pavement.
The machine shudders but stays upright. A and B both react visibly: eyes widen, shoulders recoil, then they turn to flee.
Show the normal output turning into the avalanche without a concealing cut or smoke.
(The music cuts off when the rapid stream begins.)
<Accelerating metallic clunks, mechanical rattle, cascading cans.>
4.5-8.5s — RUNNING MAKES NO DIFFERENCE
Side tracking shot. Both characters run screen-left along the same sidewalk, away from the machine. A still grips her original can in her right hand.
The rolling mass follows behind them, catches their ankles and overtakes them. A dense tumble of cans piles around their legs and torsos.
At 4.7-6.7s, A gives one continuous frightened Japanese cry {わーーーー!}, voice timbre referencing @Audio 1.
At 5.0-7.0s, B joins with {わーーーー!}, voice timbre referencing @Audio 2. Keep the two voices distinct even while overlapping.
They lose their footing and sink into the heap; keep their faces visible.
By 8.5s, both are immobilized up to their shoulders, side by side. Their voices have stopped.
Progress from running to overtaking to burial; no running loop.
8.5-10.5s — SILENCE AFTER THE AVALANCHE
Front medium two-shot. A is on the left, B on the right. Both heads and their raised hands emerge from the SAME settled heap of cans.
The machine's output stops. A few final cans roll down the pile; the metallic roar quickly fades.
A holds her original unopened can just above the heap and stares in disbelief.
B takes ONE loose can from the pile. Bodies remain naturally concealed, with coherent shoulders and arms.
<The last few cans clatter and settle, then faint street ambience.>
10.5-12.5s — THE TINY OPENING SOUND
Cut to a medium close-up of B, her face and the can held just below her chin.
Still buried, B calmly opens the pull tab once with a brief hand movement and synchronized sound.
<One crisp can-opening click and a short "psst.">
The "psst" is escaping gas, not a vocalization.
B's mouth stays closed. She looks quietly satisfied and holds the open can. No speech, drinking sequence or additional action.
Keep the pile visible around her shoulders. No music or comedy sting.
12.5-15.0s — THE FAMILIAR FACE
Cut to A's face, with cans around her shoulders and her original unopened can still visible at the lower frame edge.
Her familiar stunned, uncomprehending expression: wide eyes, small pupils, raised eyebrows, slightly parted lips held still.
Her gaze shifts once from the single can to the mountain around her, then settles forward.
Use a short controlled push-in and hold the final full second. No dialogue, scream or repeated mouth movements.
End directly on A's silent face.
OUTPUT RULES
Only A's opening sentence and the TWO assigned sustained cries. Each uses her own voice reference. No additional dialogue, narration, laughter, improvised vocal reactions or voice from the vending machine.
No subtitles, captions, speech bubbles, readable text, logos, watermark, HUD or end card.
No extra people, duplicate characters, mixed identities, extra limbs, fused hands, costume changes, injuries or disappearing can pile.
Readable progression: purchase → overflow → escape → burial → tiny opening sound → stunned face. Each shot adds a development; no repetitive action loops.
中文译文
Seedance 2.5 提示词
以所选画幅比例,创建一个全新的 15 秒日语动作喜剧短片,要求对白与音效同步。
【故事】
A 买了一罐汽水。自动贩卖机突然喷涌出数百个易拉罐。A 与 B 仓皇逃跑,却被翻滚而来的罐头雪崩追及,埋至肩部。B 从容地打开一罐。结尾落在 A 那张熟悉的惊愕脸上,手中仍握着她买的那唯一一罐。
【参考】
A = 来自 @Image 1 的黑发角色;音色来自 @Audio 1。
B = 来自 @Image 2 的白发角色;音色来自 @Audio 2。
须精确保留五官、发型、服装、配饰、身材比例与参考美术风格。参考仅定义身份与音色,而非姿势、背景或台词。不得融合身份,也不得回放已录制的词句。
【场景与风格】
黄昏时分的涩谷十字路口旁空旷人行道。一台与人等高的自动贩卖机置于画面右侧,其下方取货口朝向镜头。A 立于机器旁;B 在她左侧等候。付款已经完成。
涩谷建筑、十字路口斑马线、地面、机器、易拉罐与光照均为照片级真实质感。A 与 B 保持手绘二维动画风格,线条干净,使用赛璐璐上色,并融入接触阴影与反射光。
成排的普通饮料罐与选货按钮;不可出现可辨认的标签或标志。
【道具与动作连续性】
所有易拉罐均为手掌大小的金属圆筒,银色配青色色带与拉环。不得出现液体喷涌、泡沫、巨型罐体或无定形金属块。
每一只新出现的易拉罐都必须从同一个取货口涌出。机器保持原有尺寸与外观。
易拉罐以细窄的流束喷出,呈扇形散开,弹跳并滚落成地面雪崩,将二人吞没。
A 用右手接住第一只罐,并在最后一帧前始终紧握这只未开启的罐。她不得再次按下按钮。
仅限两名角色;二人均被埋住。须保持面部可见、身体无伤。各镜头之间须保留已堆积的罐堆。
【0.0–2.0 秒——只买一罐】
以中景双人镜头开场,呈现两位角色、选货按钮与取货口。
A 用左手按下其中一个按钮。在 0.2–1.1 秒期间,A 用日语说 {いっぽんだけ。},音色参考 @Audio 1,嘴唇自然同步。
一只易拉罐以普通的“咣当”声落入出货托盘。A 用右手将其拿起。B 在一旁随意观看。
须在 2.0 秒前完成购买动作;不出现走近或投币过程。
(轻快、短促的日常器乐背景音乐。)
<按钮点击声,一只易拉罐的普通咣当声。>
【2.0–4.5 秒——机器停不下来】
保持清晰的三分之三角度视野,纳入机器、两位角色与地面。
A 刚取走她的罐,另一只即落下,随后接连掉落更多。咣当声逐渐加速,演变为急促的机械哒哒声。
至约 2.8 秒,一股持续的易拉罐流从取货口喷涌而出,呈扇形散开,在地面上化作翻滚的雪崩。
机器颤抖但保持直立。A 与 B 均有明显反应:双目睁大、双肩一缩,随后转身奔逃。
须直接呈现正常出货转为雪崩的过程,不得用切换或烟雾进行遮挡。
(急促流束开始时,背景音乐立即切断。)
<加速的金属咣当声、机械哒哒声、罐体倾泻声。>
【4.5–8.5 秒——逃跑无济于事】
侧面跟踪镜头。两位角色沿同一条人行道向画面左侧奔跑,远离机器。A 的右手仍紧握最初那只罐。
翻滚的罐堆紧随其后,缠住她们脚踝并将其吞没。密集翻滚的罐体堆积至她们的腿部和躯干周围。
在 4.7–6.7 秒期间,A 发出一声持续的日语惊叫 {わーーーー!},音色参考 @Audio 1。
在 5.0–7.0 秒期间,B 加入一声 {わーーーー!},音色参考 @Audio 2。即使声音重叠,也须保持两种音色彼此分明。
二人失去平衡,沉入罐堆之中;保持面部可见。
至 8.5 秒,二人并肩被埋至肩部,无法动弹。喊声停止。
须呈现由奔逃、被追上到被埋的递进过程;不得出现奔跑循环。
【8.5–10.5 秒——雪崩后的沉寂】
正面中景双人镜头。A 在左,B 在右。两人的头与举起的双手从同一堆已稳定的易拉罐中露出。
机器停止出货。几只最后的罐体沿罐堆滚落;金属轰鸣声迅速消退。
A 将最初那只未开启的罐举在罐堆上方,难以置信地凝视。
B 从罐堆中取出一只松散的罐。身体自然隐没于罐中,肩部与手臂保持连贯一致。
<最后几只易拉罐叮当作响并稳定下来,随后是隐约的街景环境音。>
【10.5–12.5 秒——那声轻微的开罐声】
切至 B 的中近景镜头,包含她的脸与举在下颌附近的易拉罐。
B 仍被埋住,以一个简短的手部动作从容地一次性拉开拉环,并配有同步音效。
<一声清脆的开罐“咔哒”声与一段短促的“嘶——”声。>
“嘶——”声为气体逸出的声音,而非人声。
B 双唇紧闭。她神情安静满足,并握着已开启的罐。不得出现台词、喝饮动作或其他附加行为。
须保持其肩部周围的罐堆可见。不得使用音乐或喜剧配乐音效。
【12.5–15.0 秒——那张熟悉的脸】
切至 A 的面部镜头,肩部四周环绕易拉罐,画面下沿边仍可见她最初那只未开启的罐。
呈现她标志性的、无法理解的惊愕神态:眼睛睁大、双眉上抬、瞳孔缩小、双唇轻启并保持不动。
她的视线由那只孤零零的罐移向周围的“罐山”一次,随后定格向前。
使用一次短促克制的推镜头,并保持最后一秒定格。不得有台词、尖叫或重复的嘴部动作。
画面直接结束在 A 沉默的脸上。
【输出规则】
仅可出现 A 的那句开场台词与两句指定的持续喊声。各角色使用其对应的音色参考。不得加入额外的台词、旁白、笑声、即兴的人声反应或来自自动贩卖机的声音。
不得出现字幕、说明文字、对话气泡、可辨认文字、标志、水印、HUD 及片尾画面。
不得出现额外人物、重复角色、身份混杂、多余肢体、手部融合、服装变化、受伤或消失的罐堆。
可读的剧情递进:购买 → 溢出 → 逃跑 → 埋住 → 轻微的开罐声 → 惊愕的脸。每个镜头都须带来新进展;不得出现重复的动作循环。
来源与署名
原文由 @tokyo_Valentine 发布在 X:查看原推文。 本页逐字保留原文并提供机器翻译的中文解读;版权归原作者所有。
同工具的更多提示词
- 电影级 16:9 黑暗奇幻风格,照片写实,哥特式大教堂,翠绿色发光魔法,戏剧性低光照明,胶片颗粒感。分 3 部分生成,每部分约 10 秒。 第 1 部分 场景…@itsSaira_1
- 制作一段真实感的电影级生活方式 Vlog 视频,全程出现同一位韩国年轻女性,严格保持面部特征、发型、身材比例和形象的一致性。开场于一个明亮温馨的卧室中,她穿着…@ZorviaLux
- 阳光明媚的森林溪流,皮克斯风格3D动画。一只黄色小鸭,头上有一撮棕色绒毛,眼睛大大的,脚是橙色的,独自在清澈的水中划水,水下是卵石,周围有蕨类和覆满青苔的岩石…@Zyrellix
- 主体:全程保持完全一致的面部、身份、发型、肤色、身材比例和参考造型。仅在搭配造型的场景中更换上衣。 真实韩国女性,温馨现代的韩式公寓,早晨 vlog 美学…@doctorwasif
- 创作一段简短、超写实的日本高中恐怖动作场面,主角为一名女学生。她正走过安静的学校走廊时,一名男学生突然在她面前僵住不动。黑色的血管迅速蔓延至他的脖子、脸部和手…@AIwithSynthia
- 一个电影感的俯视顶拍镜头:一位年轻的亚洲女性赤脚躺在一块优雅的复古波斯地毯上,周围是温暖、舒适的房间,前景是缓缓旋转的吊扇木叶片。她拥有自然的黑色头发和一种平…@Zoyavelle