免费可复制的 AI 视频提示词:一份 2026 年即拿即用的合集
要点速览
- 免费可复制的 AI 视频提示词是可直接运行的模板,你可以直接从本页复制——无需注册、没有邮箱墙:复制、粘贴、替换
[方括号]、生成。 - 本合集按模型组织——Seedance 2.5、Kling 3.0、Veo 3.1 和 Wan 3.0 各六条提示词——因为各模型的”方言”差异比场景差异更大。
- 改造只需约 30 秒:用同类细节填满每个方括号,保留镜头与光线描述行,然后从表格中追加该模型的一行式附加指令。
- 任何模板都可搭配下文的通用负面块;只有当某个模型以同样方式失败两次时,才添加该模型专属的负面提示。
- 全部 24 条提示词均为本文章原创——没有任何内容抄自提示词分享网站;社区素材请见我们的免费 AI 视频提示词库指南。
- 免费指的是作为起点免费:商用前请先改造,并查看你所用生成器的服务条款——输出内容的所有权因平台而异。
- 如果你已在使用我们按场景组织的即拷即贴提示词包,那么本合集就是它的按模型组织的姊妹篇:相同的方括号约定,不同的切分维度。
什么是免费可复制的 AI 视频提示词?
免费可复制的 AI 视频提示词是预先写好的提示词模板,发布出来供任何人取用、改造并用于生成——无需注册、无需许可证密钥、无需署名。它的价值在于结构:一个可用的模板已经编码了排列顺序(主体 → 动作 → 环境 → 镜头 → 光线 → 风格,外加声音)、镜头词汇,以及初学者最常弄错的模型专属语法——你的第一次生成从基线出发,而不是从空白框开始。
本合集收录 24 条这样的提示词——每个模型六条——外加一个通用负面块和四个模型专属负面提示。
使用条款(简短而诚实)
- 作为起点免费。 可复制、修改并将这些模板用于任何目的——个人、内部、商业。无需署名,无需付费。
- 发布前先改造。 逐字照搬的模板适合测试;用于广告、商品页和盈利性帖子时,请替换方括号——一成不变的模板在规模化使用时显得千篇一律,竞争对手也能抄走同样的内容。
- 查看平台条款。 提示词自由不等于输出所有权:每个生成器的服务条款管辖你生成的内容,权利因套餐而异。商标与肖像规则同样适用于你填入的方括号。
- 我们不承诺什么。 我们不保证结果完全一致——生成器是非确定性的,厂商会发布改变行为的更新。请定期重新测试。
尽管拿去用,把它们变成你的,同时留意你生成所用的平台。
30 秒完成改造
方括号约定
[大写方括号] 中的任何内容都是可替换的槽位,并附有可用的示例——[a rain-slicked harbor at 5am]——可原样使用。用同*类*细节替换它——地点换地点、镜头换镜头——并且不要留下任何未填的槽位:未填的方括号会迫使模型自行猜测。
各模型附加指令
一个场景,四种方言:方括号填一次,附加指令追加一次。下表是本合集的后缀约定,而非厂商文档——但它反映了各方言分化程度之深:某个模型的推荐措辞会主动违反另一个模型的规则,正如 MStudio 提示词指南所记录的那样。
| 模型 | 附加指令 | 作用 | 如何追加 |
|---|---|---|---|
| Seedance 2.5 | +Sound: 行、@ 标签 |
环境音或对白作为独立图层;@ 锚定一个名词 |
末行 +Sound: [ambience];@tag 紧跟其名词之后 |
| Kling 3.0 | +Atmosphere:、有序节拍 |
情绪与场景事实分离;先 / 然后 / 最后 使序列连贯 | 末行为 +Atmosphere: [mood];把多动作拆成节拍 |
| Veo 3.1 | +Lens:、+Audio: |
镜头与音频是一等公民——Veo 原生生成音频 | 末尾行 +Lens: [camera move]、+Audio: [sound design] |
| Wan 3.0 | 内联风格关键词 | Wan 依靠句内关键词引导,而非独立图层 | 融入风格槽位:…documentary style, slight film grain |
cooly.ai 实用指南说得直白:六个核心要素中漏掉任意两个,输出几乎肯定需要重roll。这些模板保证六项全部填齐。
合集正文:24 条提示词,按模型组织
填好方括号,追加该模型的附加指令,粘贴。编号为 1–24。
Seedance 2.5(模板 1–6)
该模型想要什么: Seedance 奖励紧凑、节拍驱动的提示词——单个场景控制在约 250 词以内(超过则尾部节拍会被丢弃),把多动作镜头拆成*先 / 然后 / 最后*的节拍,给音频单独的 +Sound: 图层而不是埋在叙述里;@ 标签锚定持续存在的名词。四者中语法容忍度最高,对臃肿最不宽容。深入讲解:Seedance 2.5 提示词指南。
叙事 ×2
Template 1 (Train window): First, [a night train pulls out of a small mountain station]. Then, [rain streaks across the window as tunnel lights strobe past a sleeping passenger's reflection]. Finally, [the train bursts into a valley of dawn mist]. Interior handheld, 35mm, carriage sway; platform lamps warming to pale daylight. Naturalistic film style, light grain. +Sound: rail clatter, a distant horn.
Template 2 (Market morning): First, [a fishmonger flips open canvas awnings in a wet seaside market]. Then, [ice glints under a hanging bulb as a knife parts a silver mackerel]. Finally, the camera settles on [regulars trading coins and paper bags]. Slow lateral track, 35mm, eye level. Cool overcast light, warm stall practicals. Documentary realism. +Sound: shouted orders, knife on board, gulls.
产品 ×2
Template 3 (Sneaker turntable): A [white-and-sage running sneaker] rotates evenly on [a brushed-concrete pedestal], @shoe laces catching the light as the heel passes the lens. Locked-off medium shot, 85mm, shallow depth. Soft key from upper left, cool rim from behind, faint floor reflection. Clean catalog style, no text. +Sound: a turntable hum, one soft synth swell.
Template 4 (Espresso macro): Extreme close-up of [espresso pouring into a small ceramic cup], crema forming in ribbons, one droplet hanging and falling, @cup catching a warm highlight. Macro lens, slow tilt with the stream, razor-thin focus. Hard raking light, dark background, glossy highlights. Premium coffee-ad style. +Sound: the pour, a low pump hum.
人物 ×2
Template 5 (Barista talk): [A woman in her 30s, denim apron] leans on [a sunlit café counter] and speaks to camera with small natural gestures and steady eye contact; then she glances down as [steam curls off a fresh pour]. Static medium close-up, 85mm, shallow depth. Soft window key from camera left, warm practicals behind. Interview realism. +Sound: her voice, café murmur.
Template 6 (Street musician entrance): First, [a cellist crosses a subway mezzanine with a cello case]. Then, she sets up against [a tiled column] and draws the bow while [commuters slow into a half-circle]. Finally, the camera pulls wide as [afternoon light falls through the ceiling grates]. Tracking into a pull-back, 35mm. Mixed daylight and fluorescent. +Sound: the first long note, footsteps thinning.
Kling 3.0(模板 7–12)
该模型想要什么: Kling 依靠显式的时序编排和分离的情绪:用有序节拍陈述*发生了什么*,用专门的 +Atmosphere: 行描述*感觉如何*——把情绪揉进场景事实会让两者都变得浑浊。它用先/然后/最后的节拍和一个你应当实际使用的专属负面字段,把多镜头动作连贯地串联起来;它的口碑在人物表现:Kling 3.0 角色一致性指南。
Template 7 (Harbor dawn): First, [orange deck lamps burn as trawlers knock against a foggy pier]. Then, [a fisherman coils a wet rope as the fog thins]. Finally, [the sun clears the breakwater and the water turns to hammered copper]. Slow push-in from wide to medium, 35mm, tripod-steady. Golden backlight through fog, cool blue shadows. Muted teal-and-amber grade. +Atmosphere: quiet, cold, hopeful.
Template 8 (Forest trail): First, [sun shafts break through pines after rain]. Then, [a hiker steps over a mossy root while mist drifts between trunks]. Finally, the trail opens onto [a granite overlook above a sea of cloud]. Forward tracking from behind, 35mm, gentle handheld. Dappled backlight, moisture catching light. Travel-documentary style. +Atmosphere: serene, cool, unhurried.
Template 9 (Skincare lifestyle): First, [a glass dropper lifts from an amber serum bottle on a wet marble vanity]. Then, [a woman dots serum along her cheekbone in a mirror]. Finally, [she turns toward the window with a relaxed half-smile]. Slow push-in, 50mm, focus racking from bottle to face. Diffused window light, dewy highlights, pale palette. +Atmosphere: fresh, calm, luminous.
Template 10 (Watch macro): First, [a brushed-steel dive watch surfaces from shadow on black velvet]. Then, [the bezel rotates one detent as a light band sweeps across the sapphire crystal]. Finally, [the crown catches a hard specular flash]. Locked-off macro, 100mm. Hard key from camera right, black background, cool rim. Precision product-ad style, no text. +Atmosphere: exact, cold, premium.
Template 11 (Dancer rehearsal): First, [a contemporary dancer stretches at a barre in an empty daylit studio]. Then, [she crosses the floor in one continuous turn, skirt lagging the motion]. Finally, [she lands facing the mirror as dust motes hang in the light]. Lateral tracking, 50mm, smooth gimbal. Tall window light, grey walls, one warm bounce. +Atmosphere: weightless, intimate.
Template 12 (Chef plating): First, [a chef spoons dark reduction onto a warm plate in a steel kitchen]. Then, [tweezers place a single herb leaf as steam rises off the seared scallop]. Finally, [a waiter's hand lifts the plate out of frame]. Close medium shot, 50mm, slow push-in. Hard overhead light with warm heat-lamp fill. +Atmosphere: precise, hushed, anticipatory.
Veo 3.1(模板 13–18)
该模型想要什么: Veo 把镜头语言和音频视为一等公民——给镜头运动单独的 +Lens: 行,给声音设计单独的 +Audio: 行:Veo 原生生成音频,所以只有环境音的提示词是在浪费它最强的特性。它能吸收比 Seedance 更长、更富电影感的描述,因此以下模板更为丰满。深入阅读:Veo 3.1 原生音频与 4K 指南。
Template 13 (Crosswalk symphony): [A dense crosswalk in rain-slicked Shinjuku at blue hour], hundreds of umbrellas surging on the walk signal, neon smeared across the asphalt, a lone figure in [a yellow raincoat] crossing against the flow. Elevated long-lens wide compressing the crowd, slow tilt down as the signal changes. Cool twilight, hot neon accents. +Lens: slow tilt, telephoto compression. +Audio: umbrella patter, signal chime, crowd murmur.
Template 14 (Rooftop golden hour): [Two friends lean on a rooftop railing above a sprawling city], passing a bottle as [the sun drops between distant towers] and window lights flicker on block by block. Slow arc around the pair, 50mm, gimbal-steady. Backlit rim, soft bounce fill, flare grazing the frame. Warm indie-drama grade. +Lens: 180-degree arc, focus on the pair. +Audio: wind, traffic, easy laughter.
Template 15 (Speaker hero): A [charcoal fabric speaker] sits on [wet slate after a storm], [concentric ripples spreading from its base] as if the bass moved the water. Locked-off low-angle medium, 85mm, shallow depth. Hard key camera-left carving the fabric texture, cool sky fill, mirror reflection below. Moody tech-ad style, desaturated. +Lens: rack focus from ripple to grille. +Audio: sub-bass pulse, water lapping.
Template 16 (Perfume in motion): [A clear glass perfume bottle tumbles in slow motion through fine mist], light refracting into soft caustics on [a pale stone backdrop], the cap separating and re-aligning mid-frame. High-speed macro, 100mm, static camera. Large soft source behind glowing through the glass, cool shadows. Luxury fragrance-ad style. +Lens: static macro, 120fps. +Audio: soft whoosh, glass chime.
Template 17 (Founder interview): [A woman in a dark blazer, mid-40s] speaks to an off-camera interviewer in [a converted loft office], prototypes blurred behind her, pausing mid-thought before finishing with a small nod. Medium shot, 85mm, rule-of-thirds, shallow depth. Soft key from camera left with negative fill right, practicals as bokeh. +Lens: imperceptible push-in. +Audio: her voice, quiet room tone.
Template 18 (Runner greeting): [A runner in a red windbreaker crests a foggy bridge at sunrise], slows to a walk breathing hard, then waves to [a cyclist passing the opposite way] who returns it. Tracking alongside, 35mm, smooth vehicle mount. Low sun burning through fog from behind, long shadow on wet pavement. Cool-to-warm sports-brand grade. +Lens: lateral tracking, gentle lead. +Audio: footfalls, breath, freewheel clicks.
Wan 3.0(模板 19–24)
该模型想要什么: Wan 最适合用句内风格关键词引导,而非追加的图层——把视觉效果作为普通短语折进风格槽位(documentary style, slight film grain, muted palette),而不是它无法解析的 + 语法。它对具体的材质词响应强烈,并配合一个独立的负面字段来排除缺陷。商用场景:Wan 3.0 商业广告指南。
Template 19 (Snow village): [Lantern-lit wooden houses stacked up a snowy hillside at dusk], [smoke rising straight from tile roofs], a stone stair threading between eaves as [a figure in a dark coat climbs toward the top]. Slow tilt-up following the climb, 35mm, tripod-steady. Blue snow ambience, warm window pools, footprints holding shadow. Quiet realism, slight grain, muted palette. Sound: muffled snow steps, far-off chimes.
Template 20 (Underwater drift): [A green sea turtle glides over a shallow coral garden], [sun caustics rippling across its shell], small fish peeling away in silver flicks as it passes. Slow forward drift matching the turtle, 24mm, steady housing look. Bright overhead sun cutting blue water into light shafts, soft cyan cast. Nature documentary style, fine particulate. Sound: low current, bubble trails.
Template 21 (Perfume still life): [A faceted glass perfume bottle stands on rippled sand], [a veil of mist curling around its base], one beam throwing a long prismatic shadow to the right. Slow 20-degree orbit, 85mm, shallow depth. Hard low key at golden hour, warm rim on the glass, dark negative space behind. Editorial still-life style, warm grade. Sound: wind over sand, a glass tone.
Template 22 (Bicycle commute): [A steel-frame commuter bicycle leans into a wet cobblestone corner], [water spraying in a fine arc from the front tire], brick facades sliding past in soft blur. Low tracking from a lead vehicle, 35mm, background motion-blurred, frame sharp. Overcast silver light, one warm shop-sign glow, reflections in the stones. Urban lifestyle realism, slight handheld. Sound: tire hiss, freewheel clicks.
Template 23 (Painter at easel): [A painter in a paint-flecked shirt works at a tall window easel], dragging a loaded brush across [a half-finished canvas of storm clouds], then stepping back head tilted, palette at her side. Medium-wide shot, 50mm, locked off with a slight reframing drift. Soft north window light, dust in the air, walls carrying canvas color. Fine-art documentary style. Sound: brush on canvas, room tone.
Template 24 (Craftsman's hands): [An elderly watchmaker guides a tiny gear into place under a bench loupe], breath held, then exhales as [the balance wheel begins to oscillate]. Macro close-up on the hands, 100mm, focus racking from gear to ticking wheel. Single warm bench lamp from camera left, deep shadow, brass catching light. Intimate craft-film style, shallow depth. Sound: a ticking clock, one exhale.
负面提示块
通用块覆盖各类模型共有的缺陷——解剖结构漂移、闪烁、水印——只要该模型有负面字段即可使用。没有负面字段的(Gemini API 版 Veo),请反向表述——说出你*想要*什么(”a clear, stable frame”);点名缺陷反而可能召唤它。
Universal negative: blurry, low resolution, warped anatomy, extra fingers, deformed hands, morphing face, flickering, jitter, frame tearing, watermarks, text artifacts, oversaturated colors, plastic skin, duplicated subjects, distorted proportions.
模型专属负面针对该生成器的典型失败——只有在你亲眼见过两次之后才添加,且绝不要四个全部叠加:负面提示之间会互相钝化。
Seedance 2.5 negative: dropped final beat, truncated sequence, ignored +Sound layer, audio desync, face drift across beats, @tag ignored, subject swaps mid-sequence.
Kling 3.0 negative: muddy atmosphere blending into scene facts, incoherent beat order, gravity-defying props, over-smooth motion, mid-shot identity change, jittery hands, inconsistent wardrobe.
Veo 3.1 negative: audio-visual desync, dialogue lip flap, flat native sound design, over-saturated HDR look, warped reflections in glass, crowd face artifacts, camera drift off the +Lens instruction.
Wan 3.0 negative: style keywords ignored, oversaturated grade, morphing product edges, warped text-like logos, plastic skin, flickering highlights, unstable background geometry, prompt_extend contradictions.
收藏即用工作流:一个真正会被用起来的本地文件夹
两周后就找不到的提示词,等于弄丢了的提示词。把合集放在你工作的地方:
- 一个文件夹,四个文件。 创建
ai-prompts/,内含seedance.md、kling.md、veo.md、wan.md。把每个模型的六条模板连同负面提示粘贴进去,让提示词与护栏永不分离。 - 按用例打标签,而不是按模型。 你的生成器随项目而变,用例不会变。给每条模板打上
#narrative、#product或#character标签——一次搜索即可召回来自所有模型的候选。 - 给修改加版本。 当一次方括号替换修好了一次重roll,把它连同版本标记存回去(
v2 — camera slot now “slow arc”)。永远不要覆盖一个能用的提示词:更新不该抹掉有效的东西。 - 维护一份保留清单。 每次确定采用的生成记一行:
kling-t09_serum_v3 = keeper, client X。十行之后你就会知道哪些模板真正在用——其余的降级处理。 - 每季度重新测试。 厂商会悄悄发布模型更新。重跑你排名前三的保留项;如果质量发生变化,修复通常只需改一个方括号,而不是换一条新提示词。
按模型分文件夹,按用例打标签,就地做版本,保留项记清单——整套系统一张索引卡就写得下。
常见问题
1. 什么是免费可复制的 AI 视频提示词?
预先写好、发布供任何人复制并用于生成的提示词模板——无需注册,无需付费。六槽结构(主体 → 动作 → 环境 → 镜头 → 光线 → 风格,外加声音)和模型专属语法都已替你处理好;场景细节放在 [方括号] 里。填好、追加附加指令、粘贴。下面的 24 条是按生成器而非按场景组织的。
2. 这些提示词真的免费吗?有什么套路? 免费,没有套路,无需账户:直接从本页复制即可。代价在于:它们是起点——替你省掉的是空白框,不是改造。生成器是非确定性的,模型版本会变;输出质量由你的修改决定。
3. 我可以商用吗? 可以——作为起点。把它们改造成客户项目、产品视频和盈利性帖子;发布前先改造——逐字照搬的模板在规模化时显得千篇一律。然后查看你所用生成器的服务条款:输出所有权因平台而异,商标与肖像规则适用于你填入的方括号。
4. 使用它们需要账户或付费工具吗? 在你生成的地方就需要账户——那是平台的要求,不是我们的。阅读、复制和改造这些提示词不需要向我们提供任何东西:无需邮箱、无需登录、无需付费提示词库。免费生成器档位和付费档位一样好用。
5. 每个模型都需要不同的提示词吗?
需要的是不同的附加部件,而不是不同的场景:Seedance 用 +Sound 和 @ 标签,Kling 用 +Atmosphere,Veo 用 +Lens/+Audio,Wan 用内联风格关键词。这些模板已按模型预调校,因为厂商要求差异之大,以至于一条字符串很少能同时满足所有模型(MStudio),而且 Veo 到 Kling 的直接翻译常常损失质量(cooly.ai)。
6. 和你们的即拷即贴提示词包有什么区别? 是组织方式的区别,不是质量的区别。即拷即贴提示词包按*场景*排列——六组、每组五条,从电影感到节拍同步——所以你先挑镜头类型。本合集按*模型*排列——每个生成器六条——所以你先挑工具。方括号约定和编辑流程相同;选与你规划方式匹配的切分维度。
7. 在最终确定之前应该生成多少个变体? 三到五个,每个变体只改动一个方括号槽位——受控的差异才能显示是哪个槽位改变了结果;随机重写什么也说明不了。如果五个变体都没让输出变化,就从六个槽位重建:2026 自然提示词公式解释了排列顺序。
结语
空白的提示词框是你的工作流中最昂贵的部分——不是钱,而是重roll。这 24 条模板把最难学的东西前置了:要素顺序、镜头语言,以及四种日益分化的模型方言——每一节都是按照该模型所奖励的写法写的。复制,30 秒替换方括号,追加一行附加指令,用负面提示护住,跑”先测试后采用”。
把它们放进一个四文件文件夹,按用例打标签,每次修改都做版本记录,每季度重新测试——一旦变成你的,这套合集就会复利增长。按场景组织的姊妹篇:即拷即贴 AI 视频提示词包。社区素材:免费 AI 视频提示词库。结构方法:2026 自然提示词公式。分辨率适配模板:2K 视频模型提示词模板。模型深度内容:本批的2026 年最佳 Seedance 提示词库。
由 videosprompt.org 编辑团队审校 · 2026 年 10 月
分享文章