首页 / 正文

把普通绘图提示词拧成“电影镜头”:一套可复制的 Cinematic Prompt 公式(含三机位模组)

Mooko
发布于 2026-04-15 · 5分钟阅读
611 浏览
0 点赞 暴击点赞!

把普通提示词拧成“电影镜头”:一套可复制的 Cinematic Prompt 公式

你有没有这种崩溃时刻:

  • 明明写了“电影感”,出来还是像海报、像插画、像摆拍写真。
  • 画面干净是干净,但不“像一帧电影”。

问题通常不在“形容词不够多”。而在你没给模型镜头语言:镜头、镜头组、景别、构图、调色、感光质感。

下面这套方法很直接:把你的“核心画面意象”丢进去,输出一段结构严密的英文提示词,靠硬参数把质感钉死。


你要的不是“形容词”,是电影工业的描述方式 🎬

别再堆 beautiful / masterpiece / amazing 这种词了。

电影感来自这些:

  • 用什么机身拍的(ARRI / RED / Sony CineAlta)
  • 用什么镜头(定焦/变形宽银幕)
  • 焦段(35mm、广角带空间张力)
  • RAW/分辨率/色彩空间(画面“厚度”的来源)
  • 构图与景别(wide / medium / close-up、主体在画面哪)
  • 光比与氛围(低反差、荒凉、冷峻、霓虹……)

你只要把这些塞进一个固定结构,模型就会更听话。


终极电影感提示词公式(照抄就能用)

把提示词拼成 5 段,顺序别乱:

  1. Subject & Action(主体与动作)
  2. Setting & Framing(场景与构图)
  3. Camera Module(摄影机模组)
  4. Lighting, Color & Vibe(光影色彩与氛围)
  5. Parameters(后缀参数)

模板长这样(英文输出):

[Subject & Action] + [Setting & Framing] + [Camera Module] + [Lighting, Color & Vibe] + [Parameters]

你日常只要提供一句“核心画面意象”,剩下的交给这个结构去扩写。


三大摄影机模组:按气质选,不用纠结器材党

Module A:ARRI 经典写实 / 细腻叙事 / 文艺冷峻

适合:剧情片、现实题材、细节耐看、情绪克制。

Shot on ARRI ALEXA Mini, paired with Zeiss Master Prime lenses, 35mm wide angle, 3.2K ArriRaw sensor quality

Module B:RED 史诗大片 / 废土科幻 / 宽银幕张力

适合:末日、大片感、强戏剧冲突、横向拉开空间。

Shot on RED Weapon Dragon, paired with Panavision G series Anamorphic lenses, 35mm, Redcode RAW 6K, cinematic anamorphic look

Module C:Sony 极致锐利 / 赛博惊悚 / 极高反差

适合:赛博朋克、冷硬惊悚、霓虹反射、金属质感。

Shot on Sony CineAlta F65, paired with Panavision Primo Prime lenses, 35mm wide angle, Sony 16-bit Linear RAW, ultra-wide color gamut

选模组的偷懒口诀:

  • 想“像电影剧情”:选 A
  • 想“一看就大片宽银幕”:选 B
  • 想“锐到发冷的赛博惊悚”:选 C

光影、调色、氛围:这段建议固定保留

这段是你的“统一电影滤镜”,建议别改来改去:

incredibly high fidelity, cinematic color grading, cinematic lighting, deep depth of field, hyper-realistic, low contrast, desolate atmosphere

想换情绪也行,别一口气全换。你可以只改最后两三个词,比如把 desolate atmosphere 换成 tense atmospherenoir moodneon-drenched atmosphere


后缀参数:直接用默认就够了

如果你用 Midjourney,这段很实用:

--ar 16:9 --style raw
  • 16:9 更像电影画幅
  • style raw 少点“AI 自作主张的油腻滤镜”

你可以直接复制的“系统提示词”(喂给 Gemini / GPT 用)

把下面整段丢到对话的 system / instruction 里,然后你每次只发一句中文画面描述。

Role: Top-tier Hollywood cinematographer & AI prompt architect

Objective:
Convert the user's core visual idea into a tightly structured, high-impact ENGLISH AI image prompt using the Master Formula and one of the three Camera Modules.

Rules:
- Avoid vague adjectives like beautiful, masterpiece, amazing.
- Follow the formula order strictly.
- Final prompt must be 100% English.

Master Formula:
[Subject & Action] + [Setting & Framing] + [Camera Module] + [Lighting, Color & Vibe] + [Parameters]

Camera Modules (pick one):
A) Shot on ARRI ALEXA Mini, paired with Zeiss Master Prime lenses, 35mm wide angle, 3.2K ArriRaw sensor quality
B) Shot on RED Weapon Dragon, paired with Panavision G series Anamorphic lenses, 35mm, Redcode RAW 6K, cinematic anamorphic look
C) Shot on Sony CineAlta F65, paired with Panavision Primo Prime lenses, 35mm wide angle, Sony 16-bit Linear RAW, ultra-wide color gamut

Lighting, Color & Vibe (keep intact):
incredibly high fidelity, cinematic color grading, cinematic lighting, deep depth of field, hyper-realistic, low contrast, desolate atmosphere

Parameters:
--ar 16:9 --style raw

Output format:
1) Camera module choice: A/B/C + one-sentence reason
2) Final prompt in a code block

实战:把一句“核心画面意象”变成电影镜头

示例 1:文艺冷峻(选 A)

你的输入(中文随便写,越短越好):

雨夜的便利店门口,一个人抽烟

期望输出结构(示例):

  • **📸 摄影机模组选择:**A(更像剧情片的克制叙事)
  • 🎬 Prompt:
A lone man smoking under the convenience store awning, raindrops streaking down his jacket, subtle breath visible in cold air, a quiet moment of hesitation, minimal movement. Wide establishing shot from across the wet street, reflections of fluorescent signage on asphalt, subject placed on the right third, leading lines from road markings, shallow crowd presence blurred in the far background. Shot on ARRI ALEXA Mini, paired with Zeiss Master Prime lenses, 35mm wide angle, 3.2K ArriRaw sensor quality, incredibly high fidelity, cinematic color grading, cinematic lighting, deep depth of field, hyper-realistic, low contrast, desolate atmosphere --ar 16:9 --style raw

示例 2:末日大片(选 B)

你的输入:

沙漠里一辆报废公交车,车顶站着人

你希望更“宽银幕、史诗、荒凉”,那就选 B。

A solitary figure standing on the roof of a rusted abandoned bus half-buried in sand, coat whipping in strong wind, sand particles cutting across the frame, a distant storm wall approaching. Ultra-wide composition with vast negative space, low horizon line, the bus centered but slightly off-axis for tension, long shadows stretching toward camera, dunes forming layered depth. Shot on RED Weapon Dragon, paired with Panavision G series Anamorphic lenses, 35mm, Redcode RAW 6K, cinematic anamorphic look, incredibly high fidelity, cinematic color grading, cinematic lighting, deep depth of field, hyper-realistic, low contrast, desolate atmosphere --ar 16:9 --style raw

示例 3:赛博惊悚(选 C)

你的输入:

地铁站里,一个戴面罩的人回头看镜头

A masked figure turning back toward the camera at the edge of a subway platform, tense posture, gloved hand gripping a metal railing, droplets on the mask catching neon highlights. Medium-wide framing with strong vanishing point down the tunnel, subject in foreground left, harsh geometric lines from tiles and signage, commuters reduced to silhouettes, surveillance-camera vibe. Shot on Sony CineAlta F65, paired with Panavision Primo Prime lenses, 35mm wide angle, Sony 16-bit Linear RAW, ultra-wide color gamut, incredibly high fidelity, cinematic color grading, cinematic lighting, deep depth of field, hyper-realistic, low contrast, desolate atmosphere --ar 16:9 --style raw

你写“核心画面意象”的小抄(发给模型更容易出片)

用这 6 个点凑一句话就够:

  • 人/物:谁是主角
  • 动作:在做什么
  • 地点:在哪
  • 时间:白天/夜晚/黄昏
  • 天气:雨/雾/雪/沙尘
  • 情绪:紧张/孤独/压迫/庆祝

示例:

一个快递员站在天桥上,黄昏逆光,手里捏着皱掉的纸条,城市下方车流像河

这类输入特别“可电影化”。


避坑清单:很多人就死在这些细节上 ⚠️

  • 堆太多形容词:模型会乱补,画面更“像插画”。把精力放到构图、镜头、光。
  • 构图不写:缺 wide shot / close-up / leading lines / rule of thirds 这类词,电影感很难稳。
  • 景别和动作打架:你写“特写”,又写“看到整条街”,模型会懵。
  • 氛围词互相冲突:例如又“low contrast”又“extremely high contrast”,谁听谁的?删掉一个。
  • 每次都乱换参数:先用默认的 --ar 16:9 --style raw,稳定以后再玩花的。

进阶玩法:让画面更像“剧照”而不是“概念图”

想更像电影剧照,你可以在 Setting & Framing 里加两类信息:

  • 镜头调度感caught mid-turn, motion blur on passing crowd, foreground obstruction(像偷拍到的)
  • 真实世界的小瑕疵lens flares, fine film grain, tiny imperfections on skin, smudged glass reflections

注意:加一两条就行,加满会糊。


你现在就能用的“填空版 Prompt 模板”

把中文意象写好后,让模型产出下面这种结构:

[Who/what + action + micro-details].
[Where + time + weather + framing/composition + subject placement].
[Pick ONE camera module A/B/C].
incredibly high fidelity, cinematic color grading, cinematic lighting, deep depth of field, hyper-realistic, low contrast, desolate atmosphere --ar 16:9 --style raw

你给我一句“核心画面意象”也行,我可以按这套结构直接帮你拼一条可复制的电影感英文提示词。

OpenClaw
OpenClaw
木瓜AI支持养龙虾啦
木瓜AI龙虾专供API,限时领取免费tokens
可在 OpenClaw接入全球顶尖AI大模型
立即领取