视频 · Seedance 2.5
追着冰淇淋车铃声跑的夏日街巷
生成一段30秒的日式生活流手持短片:年轻女性听见冰淇淋车铃声后奔出家门,在温暖的街区中追赶、买冰淇淋并与孩子短暂玩耍。
Prompt
创作一段30秒的超照片级写实日本日常生活家庭录像,讲述一位20岁出头的可爱日本女性在家放松时听到远处冰淇淋车的铃声。她拿起钱包,穿上运动鞋,冲出家门,跑过安静的街区去追赶冰淇淋车。
呈现真实的手持家用摄像机画面,包括不完美的构图、轻微抖动、自动对焦变化、自然日光和逼真的动态模糊,不要商业电影质感。
她及时追上了车,买了一个香草草莓蛋筒,随后遇见正在玩球的邻家孩子们。她短暂地加入他们,然后沿着温暖的傍晚街道走回家,开心地吃完冰淇淋。
全程保持面孔、马尾辫、淡蓝色衬衫、米色裙子、白色运动鞋、奶油色单肩包和自然表情一致。跑步、身体平衡、包和头发的运动、自行车的惯性、球的运动、融化的冰淇淋、阴影以及材质之间的相互作用都应符合真实物理规律。
仅使用自然声音:自行车铃声、脚步声、呼吸声、孩子们的笑声、球的声音、小贩的对话、街区环境音和微风声。不要音乐或旁白。
负面提示词:不要动漫、CGI、塑料质感的皮肤、美颜滤镜、机械式动作、不可能的物理效果、漂浮物体、瞬间移动、扭曲的手、多余的手指、重复的孩子、面孔或服装变化、过度的镜头运动、商业风格、字幕、标志或水印。Create a 30-second ultra-photorealistic Japanese slice-of-life home-video following an adorable Japanese woman in her early 20s who hears the distant ice-cream cart bell while relaxing at home. She grabs her wallet, slips on her sneakers, rushes outside, and runs through a quiet neighborhood to catch the cart.
Show authentic handheld consumer-camera footage with imperfect framing, subtle shake, autofocus shifts, natural daylight, realistic motion blur and no commercial cinematic look.
She catches the cart just in time, buys a vanilla-strawberry cone, then meets neighborhood children playing with a ball. She joins them briefly before walking home through the warm late-afternoon street, happily finishing her ice cream.
Maintain consistent face, ponytail, pale-blue blouse, beige skirt, white sneakers, cream shoulder bag and natural expressions throughout. Use realistic physics for running, balance, bag and hair movement, bicycle momentum, ball motion, melting ice cream, shadows and material interaction.
Natural audio only: bicycle bell, footsteps, breathing, children laughing, ball, vendor dialogue, neighborhood ambience and light wind. No music or narration.
Negative prompt: No anime, CGI, plastic skin, beauty filters, robotic movement, impossible physics, floating objects, teleportation, distorted hands, extra fingers, duplicated children, face/outfit changes, excessive camera movement, commercial look, subtitles, logos or watermark.
使用建议
使用建议:将整段提示词用于支持较长时长与音频生成的视频模型;重点保留人物服装、发型和面部连续性,并检查奔跑、球体运动及融化冰淇淋等物理细节。







