Easy Prompt

在 X 上查看原帖

视频生成视频高难

古罗马泄露手机影像

一支 30 秒竖屏一镜到底视频,将公元 80 年的古罗马拍成粗粝不稳的手机纪实影像,依次经过市场、奴隶交易、酒馆斗殴、军团和斗兽场。

提示词正文

复制后可直接粘贴到模型或内部评测工具。

FORMAT: Vertical 9:16 smartphone video. 30 seconds of continuous handheld footage. A real modern person walking through ancient Rome, 80 AD, filming on their phone. NOT a film. NOT cinematic. NOT color graded. Raw casual smartphone footage — looks like genuinely leaked impossible footage someone reposted.

CAMERA BEHAVIOR: Handheld, walking pace the entire time — camera bobs naturally with each step, micro-shaky, never stabilized. The filmer is exploring casually, phone half-raised, glancing the camera toward whatever catches their attention — never lingering more than a few seconds on anything. Autofocus hunts twice. Framing is instinctive and imperfect — tops of buildings get cut off, horizon tilts. One brief pixelated digital zoom. No deliberate composition ever.

VISUAL TEXTURE: Compressed digital smartphone video. Slightly overexposed Mediterranean sky. Natural noise in shadows. Flat, ungraded phone-camera color in harsh midday sun. Bleached stone, dusty ochre, terracotta. Noon light, no mood lighting.

SECONDS 0–5: Opens mid-stride, no intro. Walking through a packed Roman street market — vendors shouting in Latin, stalls of bread, olives, live chickens, amphorae. People brush past close to the lens. The camera sweeps across the chaos while walking, catching fragments — a merchant weighing coins, a woman haggling, a donkey. Keeps moving.

SECONDS 5–10: The filmer turns a corner and passes a slave market — a raised wooden platform, a trader calling out in Latin, a line of men and women standing while buyers inspect them. The camera catches it in passing — a few seconds, slightly too long, the way you’d film something disturbing without stopping — then the filmer keeps walking. Crowd noise continues, nobody around reacts, it’s an ordinary sight to them.

SECONDS 10–15: Shouting erupts ahead — two men are brawling outside a tavern, wrestling into a vegetable stall, produce scattering, the vendor screaming at them. A small crowd instantly forms, laughing and jeering. The camera swings toward it while still walking, gets a shaky half-view through the gathered onlookers’ shoulders, then the filmer pushes past and moves on. The shouting fades behind.

SECONDS 15–21: The street opens up. A column of Roman legionaries marches directly across the filmer’s path — red tunics, segmented armor, shields, hobnailed sandals hammering the stone in unison. The filmer stops walking for a moment to let them pass, camera panning along the line. Loud, rhythmic, intimidating. The last soldier passes and the filmer immediately walks on.

SECONDS 21–27: The filmer rounds a final corner and the Colosseum fills the frame — brand new, one year old, blinding white travertine, massive beyond the phone’s ability to capture it. Enormous crowds stream toward its arches. Colored awnings. Vendors selling food outside. The filmer tilts the phone up while walking toward it, trying to fit the full height — fails, cuts off the top. Tries the digital zoom on the upper arches — pixelates badly — pulls back out.

SECONDS 27–30: Still walking closer, camera on the arches and the crowds funneling in. A muffled roar erupts from inside the arena — tens of thousands of voices at once. The camera instinctively jerks up toward the sound. Cuts off abruptly mid-motion. No resolution.

AUDIO: Constant walking footsteps on stone. Wind buffeting the mic. Overlapping Latin everywhere — vendors, arguments, the slave trader’s calls, the brawl crowd jeering, marching boots in unison, and finally the massive interior roar of the Colosseum crowd. No music. No narration.

PEOPLE: Historically accurate. Plain wool tunics in undyed cream, brown, rust. Real weathered Mediterranean faces. No Hollywood Romans, no togas on commoners. Nobody performs for the camera. Everyone mid-task, mid-transaction, mid-argument. A real living city.

使用场景

影视与叙事创意实验Seedance 2.5 视频生成视频模型能力评测

参考输出

成片:https://x.com/TechieBySA/status/2083864652388151426 原始来源:https://x.com/TechieBySA/status/2083864652388151426 发布者:TechieSA (@TechieBySA) 证据等级:creator-confirmed 规格:30s · 9:16 输入:text 工作流:文生视频

评分维度

检查镜头连贯、主体一致性、运镜与节奏是否符合提示词;优先评估时长、画幅、动作和场景约束是否被执行。

试用与模板

填写变量后复制,或保存到个人工作台模板。

这个模板没有变量,可直接复制使用。

用户评分

0 个评分
-

你的评分

登录后评分

评论

0

登录后评论

相关提示词