视频 · Seedance 2.5
韩国街巷小吃纪实短片
生成一支30秒写实街拍短片,记录年轻女性在韩国老街发现小吃、购买品尝并继续漫步的完整过程。
Prompt
为Seedance2.5创作一段高度写实、电影感的30秒视频,整段序列中都是同一位年轻韩国女性,面部身份、身材比例、发型、服装和自然肤质保持一致。
主体:
一位美丽的韩国女性,20岁出头,外貌自然可信。她穿着迷人的日常夏季服装:修身的短款 T 恤、时尚的高腰休闲短裤(时尚但符合日常街头穿搭的自然感),干净的黑色运动鞋,以及一条精致简约的项链。她的黑发松松地束在脑后,几缕自然发丝垂落在脸庞两侧。妆容极简,肤质真实,有细微瑕疵,眼神富有表现力,性格放松自信。
场景:
一条迷人的韩国老社区,时值温暖的傍晚午后。狭窄的居民街道,低矮的混凝土房屋,小型本地商店,停放的自行车和踏板车,电线杆,褪色的招牌,枝叶繁茂的树木,远处的行人以及真实的韩国街道细节。环境必须让人感觉真实有人居住,而非刻意布置。
视觉风格:
超写实的电影纪实影像,带有真实病毒式街头生活视频的质感。自然阳光,真实阴影,符合物理规律的反射,可信的皮肤与面料质感,细微的镜头瑕疵,真实的景深与自然运动。采用高质量现代拍摄,偶尔有手持移动,感觉像是人工操作而非人工生成。避免过度美颜滤镜、塑料皮肤、CGI 质感或过于完美的构图。
30秒序列:
0–5秒:
女子沿着狭窄的街区小路朝镜头方向随意地走来,偶尔环顾四周。微风自然地吹动她的头发和 T 恤。镜头从她稍后方跟随,然后移至她身旁。
5–10秒:
她注意到一家小小的社区小吃店,放慢脚步。她透过敞开的店面向里张望,自然地微笑并走近。捕捉细微的自发动作和真实的眼部运动。
10–16秒:
她向摊主点了一份受欢迎的韩国街头小吃。展示食物制作的特写细节:蒸汽升起,器具移动,滋滋作响的声音,以及真实的手部动作。她带着真切的好奇心观看,并与摊主短暂地笑着交谈。
16–22秒:
她接过热乎乎的小吃,因为仍然很烫,小心地吹了吹,然后咬下第一口。她的反应自然而真心快乐。捕捉她表情的短暂特写,随后是较宽的手持镜头。
22–27秒:
她突然注意到镜头正在拍摄她。她露出俏皮而惊讶的微笑,向镜头迈近一步,随意地做出手势,仿佛邀请观众跟随她。保持互动自然,不像事先安排好的。
27–30秒:
她转身继续沿着社区小巷走去。镜头跟随几步后自然地落在后方。结尾时她消失在街角,温暖的午后街道环境音持续响起。
摄影机:
自然的手持摄影,轻微的操作者移动,真实的行走运镜,偶尔的对焦调整,人物与周围环境之间的自然焦点转移,真实的运动模糊,轻微的曝光适应与不完美的取景。混合使用中距离跟拍、环境广角、自然特写和过肩镜头。避免过度的摄影机运动或人为的电影特效。
音频:
仅有真实的韩国社区环境音:脚步声,远处的交谈,鸟鸣,经过的踏板车,自行车,烹饪声,轻微车流,食物滋滋声,器具声以及自然街道噪音。加入女子与摊主之间简短自然的韩语对话,但对话保持随意、克制。无背景音乐。
真实感要求:
每个镜头都保持人物身份完全一致。不得改变面部、年龄、发型、服装,不出现重复人物、变形的手、多余手指、扭曲的食物、漂浮物体、不可能的阴影、不一致的光线、不自然的面部表情、僵硬的行走、橡胶般的动作、文字瑕疵或 CGI 般的环境。每一次互动都必须遵循现实物理规律。让整段序列看起来像在真实的韩国社区中自然拍摄的真实影像。
格式:
30秒,高质量Seedance 2.5生成,电影级写实,自然的16:9构图,细致的环境质感,可信的人物动作,镜头之间的连贯性,专业的画质,同时保留真实手持街头影像的真实感。Create a highly realistic, cinematic 30-second video for Seedance 2.5 featuring the same young Korean woman throughout the entire sequence, with consistent facial identity, body proportions, hairstyle, clothing, and natural skin texture.
SUBJECT:
A beautiful Korean woman in her early 20s with a natural, believable appearance. She wears an attractive everyday summer outfit: a fitted cropped T-shirt, stylish high-waisted casual shorts that are fashionable but natural for everyday streetwear, clean black sneakers, and a delicate simple necklace. Her black hair is loosely tied back with a few natural strands framing her face. Minimal makeup, realistic skin texture, subtle imperfections, expressive eyes and a relaxed, confident personality.
SETTING:
A charming older Korean neighborhood on a warm late-afternoon day. Narrow residential streets, low concrete houses, small local shops, parked bicycles and scooters, utility poles, faded signs, leafy trees, pedestrians in the distance and authentic Korean street details. The environment must feel genuinely lived-in rather than staged.
VISUAL STYLE:
Ultra-photorealistic cinematic documentary footage with the feeling of an authentic viral street-life video. Natural sunlight, realistic shadows, physically accurate reflections, believable skin and fabric textures, subtle lens imperfections, realistic depth of field and natural motion. High-quality modern capture with occasional handheld movement that feels human-operated, not artificially generated. No excessive beauty filter, plastic skin, CGI appearance or overly perfect composition.
30-SECOND SEQUENCE:
0–5 SEC:
The woman walks casually through the narrow neighborhood street toward the camera, occasionally glancing around. A gentle breeze moves her hair and T-shirt naturally. The camera follows her from slightly behind and then moves alongside her.
5–10 SEC:
She notices a tiny neighborhood snack shop and slows down. She looks through the open storefront, smiles naturally and walks closer. Capture small spontaneous gestures and realistic eye movement.
10–16 SEC:
She orders a popular Korean street snack from the vendor. Show close-up details of the food being prepared: steam rising, utensils moving, sizzling sounds and realistic hand movements. She watches with genuine curiosity and exchanges a short laugh with the vendor.
16–22 SEC:
She receives the hot snack, carefully blows on it because it is still hot, then takes her first bite. Her reaction is spontaneous and genuinely happy. Capture a brief close-up of her expression followed by a wider handheld shot.
22–27 SEC:
She suddenly notices the camera filming her. She gives a playful surprised smile, steps slightly toward the lens and casually gestures as if inviting the viewer to follow her. Keep the interaction natural and unscripted-looking.
27–30 SEC:
She turns and continues walking down the neighborhood lane. The camera follows for a few steps before naturally falling behind. End with her disappearing around a corner while the warm afternoon street ambience continues.
CAMERA:
Natural handheld cinematography, subtle operator movement, realistic walking camera motion, occasional focus adjustment, natural rack focus between the woman and surroundings, realistic motion blur, mild exposure adaptation and imperfect framing. Mix medium tracking shots, environmental wides, natural close-ups and over-the-shoulder shots. Avoid excessive camera movement or artificial cinematic effects.
AUDIO:
Authentic Korean neighborhood ambience only: footsteps, distant conversations, birds, passing scooters, bicycles, cooking sounds, light traffic, food sizzling, utensils and natural street noise. Include short natural Korean conversation between the woman and vendor, but keep dialogue casual and understated. NO background music.
REALISM REQUIREMENTS:
Maintain exact character identity across every shot. No face changes, age changes, hairstyle changes, clothing changes, duplicate people, warped hands, extra fingers, distorted food, floating objects, impossible shadows, inconsistent lighting, artificial facial expressions, unnatural walking, rubbery motion, text artifacts or CGI-looking environments. Every interaction must obey real-world physics. Make the entire sequence feel like genuine footage captured spontaneously in a real Korean neighborhood.
FORMAT:
30 seconds, high-quality Seedance 2.5 generation, cinematic realism, natural 16:9 composition, detailed environmental textures, believable human motion, coherent continuity between shots, professional image quality while retaining the authenticity of real handheld street footage.
使用建议
使用建议:需要支持长时序人物一致性、镜头调度与音频生成的文生视频模型;指定使用的模型为 Seedance 2.5,实际生成还需相应平台账户与服务权限,并建议分段生成后检查人物、服装和环境连续性。







