正在打开每日关注
正在打开每日关注
分享一条 15 秒竖屏超英飞驰穿行雅加达街头的电影感视频生成提示词，强调人脸一致性与真实物理。
Prompt - 15s | 9:16 | One continuous hyper-realistic cinematic shot. Using my face reference, create a 15-second live-action speed sequence on a busy Jakarta boulevard. The camera races inches above the road between moving traffic before revealing me flying horizontally beside it with perfectly consistent face, clothing, and body. Hair and clothes react naturally to extreme airflow. The camera pulls alongside as I briefly glance toward it with a confident smirk. Suddenly, I unleash an explosive burst of speed, generating a realistic atmospheric pressure wave that shakes the camera and blasts dust, paper, and debris across the road. Nearby vehicles subtly react to the shockwave. The camera chases at full speed but struggles to keep up as I accelerate again, leaving a turbulent wake and heat shimmer before disappearing into the distant horizon. End with an empty Jakarta boulevard as dust slowly settles. Style: Hyper-realistic 8K live action, 60fps, blockbuster cinematic realism, realistic physics, natural motion blur, aerodynamic cloth simulation, perfect face consistency, no teleportation, no magical effects, no body distortion, one unbroken take.
是什么
这是账号 Future Vibes AI 分享的一条可直接复用的 AI 视频生成提示词：15 秒、9:16 竖屏、一镜到底，要求以用户人脸为参考，在雅加达车流密集的大道上生成“超人式”低空飞驰的实拍感画面。
核心设计
提示词结构极具工程化色彩：先声明时长、画幅与“单一连续镜头”的硬约束；再按时间轴拆解镜头语言——贴地穿越车流、揭示主角横飞、回眸微笑、爆发加速、镜头追赶失败、尘埃落定的空镜收尾；随后用风格段锁定 8K/60fps 超写实规格，并追加负面约束（无瞬移、无魔法特效、无身体畸变、人脸服装完美一致），以压制视频模型常见的漂移与失真。
关键洞察
这条配方的价值在于把“电影分镜脚本 + 物理效果描述 + 一致性锁定 + 排除项”四层信息压缩进单段 Prompt，是典型的视频提示词配方思路：与其罗列形容词，不如约束模型的生成空间并描述可验证的物理反馈（气流、冲击波、热浪、碎片）。
边界与风险
推文未指明目标模型，实际效果高度依赖具体平台的连贯性与人脸参考能力；无成片画面佐证，配方是否真能稳定复现待验证；提示词较长，对上下文与指令遵循能力弱的模型可能部分失效。