AI视频生成提示词大全
可灵/Sora/即梦/Veo3 Prompt写法与模板

📅 2026-05-31 ✍️ 550W AI实验室 ⏱️ 阅读约10分钟
AI视频生成提示词教程封面

AI视频生成的效果好不好,80%取决于提示词(Prompt)的质量。同样的工具,高手写的提示词能生成电影级画面,新手写的却只能得到模糊的"AI味"视频。本文将系统讲解可灵AI、Sora 2、即梦Seedance、Veo3四大主流平台的提示词写法,并提供50+可直接复制使用的模板。

AI视频提示词的"五要素"结构

无论使用哪个平台,优质的视频提示词都遵循以下结构:

要素 说明 示例
主体(Subject) 谁/什么在画面中 A young Chinese woman with long black hair
动作(Action) 主体在做什么 walking confidently down a busy street
场景(Scene) 环境、光线、天气 neon-lit Tokyo street at night, rain-soaked pavement
镜头(Camera) 景别、运动方式 medium shot, slow tracking shot from the front
风格(Style) 画面质感、色调 cinematic, shallow depth of field, warm color grading
💡 核心原则:一个提示词只描述一个主要动作。"她一边打电话一边走路一边看手表"这种复合动作会让AI困惑,生成效果差。拆分为多个镜头分别生成。

四大平台提示词对比与适配

可灵AI提示词特点

可灵模板 - 都市甜宠场景:

A beautiful Chinese woman, 25 years old, long straight black hair, wearing a light blue summer dress, sitting alone at a cafe table by the window, looking at her phone with a gentle smile, afternoon golden sunlight streaming through the window, warm cozy atmosphere, medium close-up shot, slight camera push-in, cinematic lighting, shallow depth of field, 4K quality

Sora 2提示词特点

Sora模板 - 电影感开场:

Cinematic establishing shot of a massive modern Chinese city skyline at dusk, camera slowly descends from aerial view to street level, revealing a lone figure in a black coat standing at a crosswalk, city lights beginning to flicker on, volumetric fog, anamorphic lens flare, 24fps film grain, directed by Wong Kar-wai aesthetic, melancholic atmosphere

即梦/Seedance提示词特点

即梦模板 - 动作场景:

一位穿着黑色西装的年轻男子,在雨中的天台上,缓缓转身面对镜头,风吹起他的衣角和头发,背景是城市夜景的万家灯火,电影级光影,慢动作,近景到中景的缓慢拉远,氛围感强烈

Veo3提示词特点

Veo3模板 - 对话场景:

Two people sitting across from each other at a dimly lit restaurant table. The woman says "I've been waiting for you to say that" with a relieved smile. The man reaches across the table to hold her hand, looking emotional. Warm candlelight, intimate atmosphere, over-the-shoulder shot alternating between both characters, soft background music

10类常用场景提示词模板

1. 人物出场/登场

[Character description], making a dramatic entrance through [location], [specific action], [lighting], [camera movement], cinematic quality, [mood/atmosphere]

2. 情感对峙/冲突

Two characters facing each other in [location], [character A description] showing [emotion], [character B description] showing [emotion], tension in the air, [lighting style], medium two-shot, slight handheld camera movement, dramatic atmosphere

3. 环境空镜/转场

Establishing shot of [location] at [time of day], [weather/atmosphere details], [specific visual elements], slow [camera movement], [color grading style], no people, ambient sound design, cinematic wide angle

4. 追逐/动作场景

[Character description] running through [location], [specific running style], [obstacles/environment interaction], dynamic camera following from [angle], fast-paced editing feel, motion blur on background, urgent atmosphere

5. 回忆/闪回

Soft dreamy flashback sequence, [character as younger version] in [past location], [nostalgic action], warm golden color grading, slight overexposure, gentle lens flare, slow motion, ethereal atmosphere, vintage film quality

提示词优化的7个技巧

  1. 具体胜过抽象:不写"美丽的女人",写"25岁,鹅蛋脸,柳叶眉,穿白色真丝衬衫"
  2. 动词要有力:不写"走",写"stride confidently"或"stumble forward"
  3. 光线是灵魂:永远描述光线方向和质感(golden hour backlighting, harsh overhead fluorescent)
  4. 参考真实电影:加入"directed by [导演名] aesthetic"能快速定义风格
  5. 控制时长预期:5秒视频只写一个动作,10秒可以有动作+反应
  6. 负面提示词:部分平台支持"不要什么"(no text, no watermark, no blur)
  7. 迭代优化:每次只改一个变量,对比效果找到最优组合

生成后的必要处理

AI生成的视频通常带有平台水印(可灵Logo、即梦水印、Sora标识等),发布前必须去除。使用550W AI去字幕工具可以一键去除各平台AI视频水印,AI填充背景自然无痕。

相关教程:Veo3/Sora视频去水印 | 即梦/Seedance去水印

AI视频去水印 →

支持可灵/即梦/Sora/Veo3全平台 · AI填充自然无痕 · 批量处理