PageSourceSearch

https://tavoa.ai/assets/server-CJFnWyun.js

js tavoa.ai collected 2026-10-05 12:58:11 UTC 45,039 bytes, 354 lines download raw bytes

1import{i as e,t}from"./jsx-runtime-C_urxT9I.js";import"./runtime-GrEZ428t.js";import{t as n}from"./createLucideIcon-B45Jcamb.js";import{D as r,S as i,k as a}from"./index-Dbbb5cjE.js";var o=n(`calendar`,[[`path`,{d:`M8 2v4`,key:`1cmpym`}],[`path`,{d:`M16 2v4`,key:`4m81vk`}],[`rect`,{width:`18`,height:`18`,x:`3`,y:`4`,rx:`2`,key:`1hopcy`}],[`path`,{d:`M3 10h18`,key:`8toen8`}]]),s=e({default:()=>d,meta:()=>l}),c=t(),l={title:`Text, First Frame, or Reference: Which AI Video Workflow Do You Need?`,description:`Understand when to use text-to-video, image-to-video, first-and-last-frame animation, or reference media—and what each input actually controls.`,created_at:`2026-08-01`,author_name:`Video Lite Editorial`,author_image:`/logo.svg`,image:`/imgs/generated/cineo-hero-frame-1785487176800.png`};function u(e){let t={a:`a`,blockquote:`blockquote`,h2:`h2`,h3:`h3`,li:`li`,ol:`ol`,p:`p`,strong:`strong`,ul:`ul`,...i(),...e.components};return(0,c.jsxs)(c.Fragment,{children:[(0,c.jsx)(t.p,{children:`The input is not just something the model consumes. It is the contract for what the model is allowed to invent. Text gives it freedom; a first frame gives it a starting composition; reference media gives it a visual or rhythmic rule.`}),`
2`,(0,c.jsxs)(t.p,{children:[`Choose the workflow by asking one question: `,(0,c.jsx)(t.strong,{children:`what already exists that must survive the generation?`})]}),`
3`,(0,c.jsx)(t.h2,{children:`Text-to-video: when the shot does not exist yet`}),`
4`,(0,c.jsx)(t.p,{children:`Start with text when you are exploring a scene and no exact visual element needs to remain fixed. This is the fastest route to an establishing shot, an impossible environment, or several alternative directions.`}),`
5`,(0,c.jsx)(t.p,{children:`Text-to-video works best when the prompt defines one shot rather than an entire sequence:`}),`
6`,(0,c.jsxs)(t.blockquote,{children:[`
7`,(0,c.jsx)(t.p,{children:`Low aerial shot above a black-sand beach at dawn. A translucent tidal library rises slowly from the retreating water as pages flutter inside its glass walls. The camera advances in a straight line, quiet ocean ambience, pale gold horizon.`}),`
8`]}),`
9`,(0,c.jsx)(t.p,{children:`The model decides the architecture, layout, and details. That freedom is useful during exploration, but it is not the right choice when a specific character or product must match an approved design.`}),`
10`,(0,c.jsx)(t.h2,{children:`Image-to-video: when the opening frame matters`}),`
11`,(0,c.jsx)(t.p,{children:`Use a first frame when the scene already looks right and motion is the missing piece. The image anchors subject identity, wardrobe, palette, lens, and composition.`}),`
12`,(0,c.jsx)(t.p,{children:`The prompt should describe what happens after the uploaded frame:`}),`
13`,(0,c.jsxs)(t.blockquote,{children:[`
14`,(0,c.jsx)(t.p,{children:`She turns slowly toward the window as the curtain lifts in a light breeze. Her expression remains calm; the camera holds position and only the daylight changes across her face.`}),`
15`]}),`
16`,(0,c.jsx)(t.p,{children:`Avoid asking for a radical change that the duration cannot physically support. A close-up portrait cannot naturally become a city-wide aerial view in six seconds without turning into a transformation effect.`}),`
17`,(0,c.jsx)(t.h3,{children:`Add a last frame when the destination matters too`}),`
18`,(0,c.jsx)(t.p,{children:`First-and-last-frame generation is useful for product reveals, environmental transitions, and controlled before-and-after movement. The first frame establishes the departure; the last frame establishes the arrival.`}),`
19`,(0,c.jsx)(t.p,{children:`The prompt still needs to explain the path between them. “Transition smoothly” is not enough. Name the visible mechanism: the camera passes behind a column, fog fills the frame, the object unfolds, or daylight fades into neon night.`}),`
20`,(0,c.jsx)(t.h2,{children:`Reference-to-video: when the rule matters more than the first frame`}),`
21`,(0,c.jsx)(t.p,{children:`A reference image does not have to become the opening composition. It can define character identity, material, costume, palette, or art direction while the generated shot begins somewhere else.`}),`
22`,(0,c.jsx)(t.p,{children:`Reference audio and video work the same way at a different level:`}),`
23`,(0,c.jsxs)(t.ul,{children:[`
24`,(0,c.jsxs)(t.li,{children:[(0,c.jsx)(t.strong,{children:`Reference image:`}),` preserve identity, design, or visual language.`]}),`
25`,(0,c.jsxs)(t.li,{children:[(0,c.jsx)(t.strong,{children:`Reference audio:`}
25),` follow beat, timing, vocal quality, or sound texture.`]}),`
26`,(0,c.jsxs)(t.li,{children:[(0,c.jsx)(t.strong,{children:`Reference video:`}),` borrow motion cadence, choreography, or camera rhythm.`]}),`
27`]}),`
28`,(0,c.jsx)(t.p,{children:`State each role in the prompt. “Use this as reference” is vague; “keep the creature’s two horns and orange fur, but place it in a bright subway carriage” tells the model what to preserve and what it may change.`}),`
29`,(0,c.jsx)(t.h2,{children:`Do not mix workflows accidentally`}),`
30`,(0,c.jsxs)(t.p,{children:[`First-frame generation and reference-to-video solve different problems. The first frame says `,(0,c.jsx)(t.strong,{children:`start here`}),`. A reference says `,(0,c.jsx)(t.strong,{children:`remember this`}),`.`]}),`
31`,(0,c.jsx)(t.p,{children:`If you upload a first frame and then add several unrelated references, the model has to decide which composition and identity has priority. Video Lite separates these input modes so a shot begins with a clear contract. Choose the mode first, then attach only the assets that belong to it.`}),`
32`,(0,c.jsx)(t.h2,{children:`A practical decision tree`}),`
33`,(0,c.jsxs)(t.p,{children:[`Use `,(0,c.jsx)(t.strong,{children:`text-to-video`}),` if:`]}),`
34`,(0,c.jsxs)(t.ul,{children:[`
35`,(0,c.jsx)(t.li,{children:`you can describe the shot but cannot point to an existing image;`}),`
36`,(0,c.jsx)(t.li,{children:`variation is welcome;`}),`
37`,(0,c.jsx)(t.li,{children:`you are still exploring composition.`}),`
38`]}),`
39`,(0,c.jsxs)(t.p,{children:[`Use `,(0,c.jsx)(t.strong,{children:`a first frame`}),` if:`]}),`
40`,(0,c.jsxs)(t.ul,{children:[`
41`,(0,c.jsx)(t.li,{children:`the opening composition is approved;`}),`
42`,(0,c.jsx)(t.li,{children:`a face, product, or environment must stay recognizable;`}),`
43`,(0,c.jsx)(t.li,{children:`you mainly need natural movement from a still image.`}),`
44`]}),`
45`,(0,c.jsxs)(t.p,{children:[`Add `,(0,c.jsx)(t.strong,{children:`a last frame`}),` if:`]}),`
46`,(0,c.jsxs)(t.ul,{children:[`
47`,(0,c.jsx)(t.li,{children:`the shot must arrive at a specific composition;`}),`
48`,(0,c.jsx)(t.li,{children:`you are designing a transition between two known states.`}),`
49`]}),`
50`,(0,c.jsxs)(t.p,{children:[`Use `,(0,c.jsx)(t.strong,{children:`reference media`}),` if:`]}),`
51`,(0,c.jsxs)(t.ul,{children:[`
52`,(0,c.jsx)(t.li,{children:`identity, style, rhythm, or choreography must carry into a new shot;`}),`
53`,(0,c.jsx)(t.li,{children:`the generated clip should not be forced to start on the reference frame.`}),`
54`]}),`
55`,(0,c.jsx)(t.h2,{children:`Prepare references before uploading`}),`
56`,(0,c.jsx)(t.p,{children:`A strong reference is easy to read. Crop out irrelevant UI, avoid tiny subjects, and use a frame where the important identity is visible. For a character, choose a clear face and silhouette. For motion, choose a short clip whose rhythm is obvious. For audio, remove long silence before the useful beat.`}),`
57`,(0,c.jsx)(t.p,{children:`More references do not automatically mean more control. Two assets with conflicting lighting, proportions, or costumes can reduce consistency. Start with one strong reference and add another only when it supplies a different, necessary constraint.`}),`
58`,(0,c.jsx)(t.h2,{children:`Move between workflows instead of choosing only once`}),`
59`,(0,c.jsx)(t.p,{children:`The most reliable production loop uses all three modes at different stages:`}),`
60`,(0,c.jsxs)(t.ol,{children:[`
61`,(0,c.jsx)(t.li,{children:`Explore the scene with text-to-video.`}),`
62`,(0,c.jsx)(t.li,{children:`Save the strongest frame from the best direction.`}),`
63`,(0,c.jsx)(t.li,{children:`Use that frame as the start of a more controlled image-to-video shot.`}),`
64`,(0,c.jsx)(t.li,{children:`Reuse the resulting character or motion as reference material for the next shot.`}),`
65`,(0,c.jsxs)(t.li,{children:[`Keep each result in `,(0,c.jsx)(t.a,{href:`/my-creations`,children:`My Creations`}),` so the prompt and source material remain traceable.`]}),`
66`]}),`
67`,(0,c.jsxs)(t.p,{children:[`You are not choosing one permanent mode. You are deciding what the model may invent on this shot. `,(0,c.jsx)(t.a,{href:`/ai/video`,children:`Start with the generator`}),` and give every reference one clear job.`]})]})}function d(e={}){let{wrapper:t}={...i(),...e.components};return t?(0,c.jsx)(t,{...e,children:(0,c.jsx)(u,{...e})}):u(e)}var f=e({default:()=>h,meta:()=>p}),p={title:`纯文字、首帧还是参考素材:AI 视频工作流怎么选?`,description:`理解文生视频、图生视频、首尾帧动画和参考素材各自适合的场景,以及每一种输入到底控制什么。`,created_at:`2026-08-01`,author_name:`Video Lite 编辑部`,author_image:`/logo.svg`,
67image:`/imgs/generated/cineo-hero-frame-1785487176800.png`};function m(e){let t={a:`a`,blockquote:`blockquote`,h2:`h2`,h3:`h3`,li:`li`,ol:`ol`,p:`p`,strong:`strong`,ul:`ul`,...i(),...e.components};return(0,c.jsxs)(c.Fragment,{children:[(0,c.jsx)(t.p,{children:`输入素材不只是模型要读取的文件,它同时规定了“模型可以自由发明什么”。文字给它最大的空间,首帧规定开场构图,参考素材则规定身份、风格或节奏。`}),`
68`,(0,c.jsxs)(t.p,{children:[`选择工作流时,只需要先回答一个问题: `,(0,c.jsx)(t.strong,{children:`现在已经有什么东西,必须在生成后继续保留下来?`})]}),`
69`,(0,c.jsx)(t.h2,{children:`文生视频:镜头还不存在`}),`
70`,(0,c.jsx)(t.p,{children:`当你正在探索场景,并且没有任何视觉元素必须严格固定时,可以从纯文字开始。这是生成建立镜头、幻想环境和多个方向草案最快的方式。`}),`
71`,(0,c.jsx)(t.p,{children:`文生视频更适合描述一个镜头,而不是一整段剧情:`}),`
72`,(0,c.jsxs)(t.blockquote,{children:[`
73`,(0,c.jsx)(t.p,{children:`黎明时分,摄影机低空掠过黑沙海滩。退潮的海水中,一座半透明的潮汐图书馆缓慢升起,玻璃墙内的书页随风翻动。摄影机沿直线推进,安静海浪声,地平线泛着淡金色。`}),`
74`]}),`
75`,(0,c.jsx)(t.p,{children:`建筑结构、空间布局和细节都由模型决定。探索阶段需要这种自由,但当角色或产品必须符合已经通过的设计时,它就不再是最合适的选择。`}),`
76`,(0,c.jsx)(t.h2,{children:`图生视频:开场画面已经确定`}),`
77`,(0,c.jsx)(t.p,{children:`如果静态画面已经正确,只缺运动,就使用首帧。首帧会固定主体身份、服装、配色、镜头和构图。`}),`
78`,(0,c.jsx)(t.p,{children:`提示词应该从上传图片之后开始写:`}),`
79`,(0,c.jsxs)(t.blockquote,{children:[`
80`,(0,c.jsx)(t.p,{children:`她缓慢转向窗边,窗帘被微风轻轻吹起。表情保持平静,摄影机固定不动,只有日光在她脸上缓慢变化。`}),`
81`]}),`
82`,(0,c.jsx)(t.p,{children:`不要要求一个在所选时长内无法自然完成的巨大变化。一个人物特写很难在 6 秒内自然变成城市航拍,除非你本来就想要明显的变形转场。`}),`
83`,(0,c.jsx)(t.h3,{children:`终点也很重要时,加入尾帧`}),`
84`,(0,c.jsx)(t.p,{children:`首尾帧适合产品揭示、环境变化和可控的前后状态。首帧规定出发点,尾帧规定最终落点。`}),`
85`,(0,c.jsx)(t.p,{children:`提示词仍然需要解释中间怎样发生。“平滑过渡”还不够,应该写出看得见的机制,例如摄影机从柱子后穿过、雾气填满画面、物体逐层展开,或者日光逐渐变成霓虹夜景。`}),`
86`,(0,c.jsx)(t.h2,{children:`参考生视频:规则比第一帧更重要`}),`
87`,(0,c.jsx)(t.p,{children:`参考图片不一定要成为视频开场。它可以只负责角色身份、材质、服装、配色或美术方向,而生成镜头从新的构图开始。`}),`
88`,(0,c.jsx)(t.p,{children:`参考音频和参考视频只是控制了不同层面:`}),`
89`,(0,c.jsxs)(t.ul,{children:[`
90`,(0,c.jsxs)(t.li,{children:[(0,c.jsx)(t.strong,{children:`参考图片:`}),` 保持身份、设计或视觉语言。`]}),`
91`,(0,c.jsxs)(t.li,{children:[(0,c.jsx)(t.strong,{children:`参考音频:`}),` 跟随节拍、时间点、音色或声音质感。`]}),`
92`,(0,c.jsxs)(t.li,{children:[(0,c.jsx)(t.strong,{children:`参考视频:`}),` 借用动作节奏、编排或摄影机韵律。`]}),`
93`]}),`
94`,(0,c.jsx)(t.p,{children:`要在提示词里说明每份素材的职责。“参考这张图”太模糊;“保留小怪物的两只角和橙色绒毛,但把它放进明亮的地铁车厢”才说明了什么必须保留、什么允许变化。`}),`
95`,(0,c.jsx)(t.h2,{children:`不要无意中混用工作流`}),`
96`,(0,c.jsxs)(t.p,{children:[`首帧和参考生视频解决的是不同问题。首帧的意思是 `,(0,c.jsx)(t.strong,{children:`从这里开始`}),`,参考素材的意思是 `,(0,c.jsx)(t.strong,{children:`记住这个规则`}),`。`]}),`
97`,(0,c.jsx)(t.p,{children:`如果上传首帧后又加入多份不相关参考,模型就必须自己判断构图和身份谁优先。Video Lite 将两种输入方式分开,就是为了让每个镜头从明确约束开始。先选择模式,再只上传属于这个模式的素材。`}),`
98`,(0,c.jsx)(t.h2,{children:`一个实用的选择方法`}),`
99`,(0,c.jsxs)(t.p,{children:[`以下情况使用 `,(0,c.jsx)(t.strong,{children:`文生视频`}),`:`]}),`
100`,(0,c.jsxs)(t.ul,{children:[`
101`,(0,c.jsx)(t.li,{children:`能描述镜头,但没有现成图片;`}),`
102`,(0,c.jsx)(t.li,{children:`允许模型提供多种变化;`}),`
103`,(0,c.jsx)(t.li,{children:`仍在探索构图。`}),`
104`]}),`
105`,(0,c.jsxs)(t.p,{children:[`以下情况使用 `,(0,c.jsx)(t.strong,{children:`首帧`}),`:`]}),`
106`,(0,c.jsxs)(t.ul,{children:[`
107`,(0,c.jsx)(t.li,{children:`开场构图已经确认;`}),`
108`,(0,c.jsx)(t.li,{children:`人脸、产品或环境必须保持可识别;`}),`
109`,(0,c.jsx)(t.li,{children:`主要目标是让静态图片自然运动。`}),`
110`]}),`
111`,(0,c.jsxs)(t.p,{children:[`以下情况加入 `,(0,c.jsx)(t.strong,{children:`尾帧`}),`:`]}),`
112`,(0,c.jsxs)(t.ul,{children:[`
113`,(0,c.jsx)(t.li,{children:`镜头必须落到指定构图;`}),`
114`,(0,c.jsx)(t.li,{children:`正在设计两个已知状态之间的转场。`}),`
115`]}),`
116`,(0,c.jsxs)(t.p,{children:[`以下情况使用 `,(0,c.jsx)(t.strong,{children:`参考素材`}),`:`]}),`
117`,(0,c.jsxs)(t.ul,{children:[`
118`,(0,c.jsx)(t.li,{children:`身份、风格、节奏或动作编排需要进入新镜头;`}),`
119`,(0,c.jsx)(t.li,{children:`生成视频不应该被迫从参考素材的第一帧开始。`}),`
120`]}),`
121`,(0,c.jsx)(t.h2,{children:`上传前先整理参考素材`}),`
122`,(0,c.jsx)(t.p,{children:`好的参考素材应该容易读懂。裁掉无关界面,不要让主体小到看不清,并选择最能体现关键身份的一帧。角色素材要看得清脸和轮廓,动作素材要节奏明确,音频则应该删掉有效节拍之前的长时间空白。`}),`
123`,(0,c.jsx)(t.p,{children:`参考越多不等于控制越强。光线、比例或服装互相冲突的两份素材,反而会降低一致性。先使用一份强参考,只有第二份素材提供了不同且必要的限制时,再把它加入。`}),`
124`,(0,c.jsx)(t.h2,{children:`不必永远只选一种模式`}),`
125`,(0,c.jsx)(t.p,{children:`更稳定的制作流程,会在不同阶段使用三种模式:`}),`
126`,(0,c.jsxs)(t.ol,{children:[`
127`,(0,c.jsx)(t.li,{children:`用文生视频探索场景。`}),`
128`,(0,c.jsx)(t.li,{children:`从最佳方向中保存最强的一帧。`}),`
129`,(0,c.jsx)(t.li,{children:`把这张图作为首帧,生成更可控的图生视频镜头。`}),`
130`,(0,c.jsx)(t.li,{children:`再把得到的角色或运动作为下一条视频的参考素材。`}),`
131`,(0,c.jsxs)(t.li,{children:[`把结果保存在`,(0,c.jsx)(t.a,{href:`/my-creations`,children:`创作中心`}),`,让提示词和来源素材可以继续追溯。`]}),`
132`]}),`
133`,(0,c.jsxs)(t.p,{children:[`你不是在永久选择一种模式,而是在决定“这一条镜头里,模型可以发明什么”。现在可以`,(0,c.jsx)(t.a,{href:`/ai/video`,children:`打开生成器`}),`,并给每一份参考素材安排一个明确任务。`]})]})}function h(e={}){let{wrapper:t}={...i(),...e.components};return t?(0,c.jsx)(t,{...e,children:(0,c.jsx)(m,{...e})}):m(e)}var g=e({default:()=>y,meta:()=>_}),_={title:`How to Choose an AI Video Model`,description:`Choose an AI video model by shot type, duration, resolution, audio, and the way you want to direct motion.`,created_at:`2026-08-07`,author_name:`Video Lite Editorial`,author_image:`/logo.svg`,image:`/images/showcase/generated/night-conservatory.png`};function v(e){let t={a:`a`,blockquote:`blockquote`,h2:`h2`,li:`li`,ol:`ol`,p:`p`,strong:`strong`,ul:`ul`,...i(),...e.components};return(0,c.jsxs)(c.Fragment,{children:[(0,c.jsx)(t.p,{children:`The best AI video model is not the one with the longest feature list. It is the one whose controls already match the shot you want to make. Decide the delivery format, duration, and source material first; then choose from the models available in the generator.`}),`
134`,(0,c.jsx)(t.p,{children:`Video Lite supports multiple AI video models in one workflow, so you can change direction without rebuilding the prompt or moving your references to another tool.`}),`
135`,(0,c.jsx)(t.h2,{children:`Start with the shot requirements`}),`
136`,(0,c.jsx)(t.p,{children:`Before choosing a model, write down the constraints that cannot change:`}),`
137`,(0,c.jsxs)(t.ul,{children:[`
138`,(0,c.jsxs)(t.li,{children:[(0,c.jsx)(t.strong,{children:`Duration`}),` — decide whether the edit needs a fixed clip length or room to explore.`]}),`
139`,(0,c.jsxs)(t.li,{children:[(0,c.jsx)(t.strong,{children:`Resolution`}
139),` — test motion efficiently, then move to the resolution required for delivery.`]}),`
140`,(0,c.jsxs)(t.li,{children:[(0,c.jsx)(t.strong,{children:`Audio`}),` — choose an audio-capable model only when sound needs to be generated with the picture.`]}),`
141`,(0,c.jsxs)(t.li,{children:[(0,c.jsx)(t.strong,{children:`Aspect ratio`}),` — select the native format for the final channel instead of cropping afterward.`]}),`
142`,(0,c.jsxs)(t.li,{children:[(0,c.jsx)(t.strong,{children:`Reference media`}),` — check that the model supports the first frame, last frame, image, audio, or video reference you intend to use.`]}),`
143`]}),`
144`,(0,c.jsx)(t.p,{children:`The generator updates its controls for the selected model, making unsupported combinations unavailable before a render starts.`}),`
145`,(0,c.jsx)(t.h2,{children:`Match the model to the source material`}),`
146`,(0,c.jsx)(t.p,{children:`The same prompt can produce very different results depending on the input mode:`}),`
147`,(0,c.jsxs)(t.ul,{children:[`
148`,(0,c.jsxs)(t.li,{children:[(0,c.jsx)(t.strong,{children:`Text only`}),` gives the model the most freedom. Use it for exploration and establishing shots.`]}),`
149`,(0,c.jsxs)(t.li,{children:[(0,c.jsx)(t.strong,{children:`First frame`}),` locks the opening composition. Use it when a character, product, or art direction must survive the first moment.`]}),`
150`,(0,c.jsxs)(t.li,{children:[(0,c.jsx)(t.strong,{children:`First and last frames`}),` define a visual transition. Keep the change physically plausible for the selected duration.`]}),`
151`,(0,c.jsxs)(t.li,{children:[(0,c.jsx)(t.strong,{children:`Reference image, audio, or video`}),` supplies identity, rhythm, or motion language without forcing the clip to begin on that exact frame.`]}),`
152`]}),`
153`,(0,c.jsx)(t.p,{children:`Do not attach every available reference just because a model accepts it. Each reference should have one clear job.`}),`
154`,(0,c.jsx)(t.h2,{children:`Compare models with the same prompt`}),`
155`,(0,c.jsx)(t.p,{children:`When two models both fit the delivery requirements, run the same prompt and reference material through each one. Compare the qualities that matter for the shot: camera motion, subject consistency, physical realism, expression, pacing, and sound.`}),`
156`,(0,c.jsx)(t.p,{children:`Change only the model on the first comparison. If the prompt, duration, and references also change, it becomes difficult to understand which model produced the difference.`}),`
157`,(0,c.jsxs)(t.blockquote,{children:[`
158`,(0,c.jsx)(t.p,{children:`Wide shot of a glass observatory after rain. A child in a yellow coat walks slowly between glowing plants as the camera tracks at waist height. Reflections move naturally across the floor, cool blue moonlight, restrained cinematic motion.`}),`
159`]}),`
160`,(0,c.jsx)(t.p,{children:`The prompt describes one continuous event, which makes model comparisons easier to judge.`}),`
161`,(0,c.jsx)(t.h2,{children:`A workflow that wastes fewer generations`}),`
162`,(0,c.jsxs)(t.ol,{children:[`
163`,(0,c.jsx)(t.li,{children:`Write the shot as a single visible action.`}),`
164`,(0,c.jsx)(t.li,{children:`Choose a model that supports the required input and output format.`}),`
165`,(0,c.jsx)(t.li,{children:`Test the composition at the lower practical resolution.`}),`
166`,(0,c.jsx)(t.li,{children:`Change only one variable between attempts: camera, subject motion, or timing.`}),`
167`,(0,c.jsx)(t.li,{children:`Add reference media only when identity, composition, or motion begins to drift.`}),`
168`,(0,c.jsx)(t.li,{children:`Move to the final resolution after the motion works.`}),`
169`]}),`
170`,(0,c.jsxs)(t.p,{children:[`The model should follow the production decision, not replace it. `,(0,c.jsx)(t.a,{href:`/ai/video`,children:`Open the generator`}),`, compare the available models, and direct one shot at a time.`]})]})}function y(e={}){let{wrapper:t}={...i(),...e.components};return t?(0,c.jsx)(t,{...e,children:(0,c.jsx)(v,{...e})}):v(e)}var b=e({default:()=>C,meta:()=>x}),x={title:`如何选择 AI 视频模型`,description:`根据镜头类型、时长、清晰度、音频和运动控制方式选择合适的 AI 视频模型。`,created_at:`2026-08-07`,author_name:`Video Lite 编辑部`,author_image:`/logo.svg`,
170image:`/images/showcase/generated/night-conservatory.png`};function S(e){let t={a:`a`,blockquote:`blockquote`,h2:`h2`,li:`li`,ol:`ol`,p:`p`,strong:`strong`,ul:`ul`,...i(),...e.components};return(0,c.jsxs)(c.Fragment,{children:[(0,c.jsx)(t.p,{children:`最合适的 AI 视频模型,不是参数列表最长的那个,而是控制项恰好符合当前镜头的那个。先决定交付比例、时长和参考素材,再从生成器支持的模型中选择,会比先选模型再凑提示词稳定得多。`}),`
171`,(0,c.jsx)(t.p,{children:`Video Lite 在同一套工作流中支持多个 AI 视频模型,切换方向时不需要重新编写提示词,也不用把参考素材迁移到其他工具。`}),`
172`,(0,c.jsx)(t.h2,{children:`先确定镜头要求`}),`
173`,(0,c.jsx)(t.p,{children:`选择模型前,先写下不能改变的条件:`}),`
174`,(0,c.jsxs)(t.ul,{children:[`
175`,(0,c.jsxs)(t.li,{children:[(0,c.jsx)(t.strong,{children:`时长`}),`:明确剪辑需要固定时长,还是需要更灵活的探索空间。`]}),`
176`,(0,c.jsxs)(t.li,{children:[(0,c.jsx)(t.strong,{children:`清晰度`}),`:先高效测试运动,再切换到交付所需的清晰度。`]}),`
177`,(0,c.jsxs)(t.li,{children:[(0,c.jsx)(t.strong,{children:`音频`}),`:只有声音需要与画面同时生成时,才选择支持音频的模型。`]}),`
178`,(0,c.jsxs)(t.li,{children:[(0,c.jsx)(t.strong,{children:`画幅`}),`:直接选择最终发布渠道需要的原生比例,避免生成后裁切。`]}),`
179`,(0,c.jsxs)(t.li,{children:[(0,c.jsx)(t.strong,{children:`参考素材`}),`:确认模型支持你准备使用的首帧、尾帧、图片、音频或视频参考。`]}),`
180`]}),`
181`,(0,c.jsx)(t.p,{children:`生成器会根据所选模型自动更新控制项,在渲染前排除不支持的参数组合。`}),`
182`,(0,c.jsx)(t.h2,{children:`根据素材输入选择模型`}),`
183`,(0,c.jsx)(t.p,{children:`同一条提示词使用不同输入方式时,结果也会明显不同:`}),`
184`,(0,c.jsxs)(t.ul,{children:[`
185`,(0,c.jsxs)(t.li,{children:[(0,c.jsx)(t.strong,{children:`纯文字`}),` 给模型最大的创作空间,适合探索方向和建立镜头。`]}),`
186`,(0,c.jsxs)(t.li,{children:[(0,c.jsx)(t.strong,{children:`首帧图片`}),` 固定开场构图,适合需要保持角色、产品或美术方向的场景。`]}),`
187`,(0,c.jsxs)(t.li,{children:[(0,c.jsx)(t.strong,{children:`首帧与尾帧`}),` 定义视觉转变,但变化必须在所选时长内符合物理逻辑。`]}),`
188`,(0,c.jsxs)(t.li,{children:[(0,c.jsx)(t.strong,{children:`参考图片、音频或视频`}),` 提供身份、节奏或运动语言,不要求生成结果从参考素材的第一帧开始。`]}),`
189`]}),`
190`,(0,c.jsx)(t.p,{children:`不要因为模型支持,就一次挂上所有素材。每一份参考素材都应该有一个明确任务。`}),`
191`,(0,c.jsx)(t.h2,{children:`使用同一条提示词比较模型`}),`
192`,(0,c.jsx)(t.p,{children:`当多个模型都符合交付要求时,可以使用相同的提示词和参考素材分别生成,再比较镜头真正需要的特质:运镜、主体一致性、物理真实感、表情、节奏和声音。`}),`
193`,(0,c.jsx)(t.p,{children:`第一次比较时只切换模型。假如提示词、时长和参考素材也同时变化,就很难判断差异究竟来自哪里。`}),`
194`,(0,c.jsxs)(t.blockquote,{children:[`
195`,(0,c.jsx)(t.p,{children:`雨后的玻璃温室广角镜头。一位穿黄色雨衣的孩子缓慢走过发光植物,摄影机以腰部高度平稳跟拍。地面反射自然移动,冷蓝色月光,克制的电影感运动。`}),`
196`]}),`
197`,(0,c.jsx)(t.p,{children:`这条提示词只描述一个连续事件,更容易判断不同模型的表现差异。`}),`
198`,(0,c.jsx)(t.h2,{children:`更省生成次数的工作流`}),`
199`,(0,c.jsxs)(t.ol,{children:[`
200`,(0,c.jsx)(t.li,{children:`把镜头写成一个可以看见的连续动作。`}),`
201`,(0,c.jsx)(t.li,{children:`选择支持所需输入方式和交付格式的模型。`}),`
202`,(0,c.jsx)(t.li,{children:`先用较低但可判断的清晰度测试构图。`}),`
203`,(0,c.jsx)(t.li,{children:`每次只改一个变量:摄影机、主体动作或节奏。`}),`
204`,(0,c.jsx)(t.li,{children:`只有身份、构图或运动开始漂移时,再加入参考素材。`}),`
205`,(0,c.jsx)(t.li,{children:`运动确定之后,最后再切换到交付清晰度。`}),`
206`]}),`
207`,(0,c.jsxs)(t.p,{children:[`模型应该服从制作决策,而不是代替制作决策。现在可以`,(0,c.jsx)(t.a,{href:`/ai/video`,children:`打开生成器`}),`,比较所有可用模型,再一次导演一个镜头。`]})]})}function C(e={}){let{wrapper:t}={...i(),...e.components};return t?(0,c.jsx)(t,{...e,children:(0,c.jsx)(S,{...e})}):S(e)}var w=e({default:()=>D,meta:()=>T}),T={title:`How to Write AI Video Prompts That Actually Move`,description:`A practical shot-building method for describing subject motion, camera movement, timing, light, and sound without turning the prompt into a screenplay.`,created_at:`2026-08-04`,author_name:`Video Lite Editorial`,author_image:`/logo.svg`,
207image:`/images/showcase/generated/monster-bakery.png`};function E(e){let t={a:`a`,blockquote:`blockquote`,h2:`h2`,li:`li`,p:`p`,strong:`strong`,ul:`ul`,...i(),...e.components};return(0,c.jsxs)(c.Fragment,{children:[(0,c.jsx)(t.p,{children:`An image prompt describes what exists. A video prompt must also explain what changes. If the prompt contains a beautiful scene but no event, the model has to invent the motion—and invented motion is usually where drifting faces, rubber objects, and wandering cameras begin.`}),`
208`,(0,c.jsx)(t.p,{children:`The fix is not a longer paragraph. It is a clearer shot.`}),`
209`,(0,c.jsx)(t.h2,{children:`Build the prompt in five decisions`}),`
210`,(0,c.jsx)(t.p,{children:`Use this order:`}),`
211`,(0,c.jsx)(t.p,{children:(0,c.jsx)(t.strong,{children:`Subject and place → visible action → camera → light and atmosphere → timing or sound.`})}),`
212`,(0,c.jsx)(t.p,{children:`Start with a weak prompt:`}),`
213`,(0,c.jsxs)(t.blockquote,{children:[`
214`,(0,c.jsx)(t.p,{children:`A cute orange monster in a magical bakery, cinematic 3D animation.`}),`
215`]}),`
216`,(0,c.jsx)(t.p,{children:`It describes a frame, not a clip. Now direct the shot:`}),`
217`,(0,c.jsxs)(t.blockquote,{children:[`
218`,(0,c.jsx)(t.p,{children:`A small orange monster stands beside a stone oven in a moonlit bakery. A round cake suddenly rises, releasing a soft burst of flour; the monster steps back, blinks twice, then grins. The camera makes a gentle waist-height push-in. Warm oven light, cool blue window light, playful family-animation timing, six seconds, no text.`}),`
219`]}),`
220`,(0,c.jsx)(t.p,{children:`Every clause now has a job. The subject acts, the camera has one move, the lighting has two sources, and the moment has an ending.`}),`
221`,(0,c.jsx)(t.h2,{children:`Give the subject one readable action`}),`
222`,(0,c.jsx)(t.p,{children:`“Dancing, running, turning, laughing, and waving” is not richer direction. It is five actions competing for a few seconds. Choose the one action that changes the frame.`}),`
223`,(0,c.jsx)(t.p,{children:`Good actions have a visible start and finish:`}),`
224`,(0,c.jsxs)(t.ul,{children:[`
225`,(0,c.jsx)(t.li,{children:`opens an umbrella and looks up;`}),`
226`,(0,c.jsx)(t.li,{children:`steps onto the train as the doors close;`}),`
227`,(0,c.jsx)(t.li,{children:`raises a glass sphere into the sunlight;`}),`
228`,(0,c.jsx)(t.li,{children:`turns from the window and meets the camera;`}),`
229`,(0,c.jsx)(t.li,{children:`unfolds from paper into a flying dragon.`}),`
230`]}),`
231`,(0,c.jsx)(t.p,{children:`When a shot fails, simplify the action before changing the style.`}),`
232`,(0,c.jsx)(t.h2,{children:`Treat the camera like physical equipment`}),`
233`,(0,c.jsx)(t.p,{children:`Camera language works best when it can be performed by a real operator. Pick one primary move:`}),`
234`,(0,c.jsxs)(t.ul,{children:[`
235`,(0,c.jsxs)(t.li,{children:[(0,c.jsx)(t.strong,{children:`Locked-off:`}),` the camera does not move; motion comes from the subject.`]}),`
236`,(0,c.jsxs)(t.li,{children:[(0,c.jsx)(t.strong,{children:`Push-in or pull-back:`}),` changes emotional distance.`]}),`
237`,(0,c.jsxs)(t.li,{children:[(0,c.jsx)(t.strong,{children:`Track left or right:`}),` follows a subject through space.`]}),`
238`,(0,c.jsxs)(t.li,{children:[(0,c.jsx)(t.strong,{children:`Orbit:`}),` reveals shape, scale, or environment around a still subject.`]}),`
239`,(0,c.jsxs)(t.li,{children:[(0,c.jsx)(t.strong,{children:`Handheld follow:`}),` introduces energy and small natural movement.`]}),`
240`,(0,c.jsxs)(t.li,{children:[(0,c.jsx)(t.strong,{children:`Crane or aerial rise:`}),` reveals a larger world at the end of the shot.`]}),`
241`]}),`
242`,(0,c.jsx)(t.p,{children:`Avoid combining “rapid drone dive, orbit, handheld shake, and slow zoom.” Those instructions imply different rigs and different speeds.`}),`
243`,(0,c.jsx)(t.h2,{children:`Describe motion quality, not just motion type`}),`
244`,(0,c.jsxs)(t.p,{children:[`The model needs to know how the movement behaves. Useful words include `,(0,c.jsx)(t.strong,{children:`measured`}),`, `,(0,c.jsx)(t.strong,{children:`weighty`}),`, `,(0,c.jsx)(t.strong,{children:`buoyant`}),`, `,(0,c.jsx)(t.strong,{children:`hesitant`}),`, `,(0,c.jsx)(t.strong,{children:`wind-driven`}),`, `,(0,c.jsx)(t.strong,{children:`mechanical`}),`, and `,(0,c.jsx)(t.strong,{children:`barely perceptible`}),`.`]}),`
245`,(0,c.jsx)(t.p,{children:`Compare:`}),`
246`,(0,c.jsxs)(t.blockquote,{children:[`
247`,(0,c.jsx)(t.p,{children:`Flowers move in the wind.`}),`
248`]}),`
249`,(0,c.jsx)(t.p,{children:`with:`}),`
250`,(0,c.jsxs)(t.blockquote,{children:[`
251`,(0,c.jsx)(t.p,{children:`Tall poppies bend in slow uneven waves as a light desert wind passes from foreground to background; loose petals occasionally cross the lens.`}),`
252`]}),`
253`,(0,c.jsx)(t.p,{children:`The second version gives the motion direction, rhythm, and depth.`}),`
254`,(0,c.jsx)(t.h2,{children:`Use references to remove description, not add more`}),`
255`,(0,c.jsx)(t.p,{children:`If you upload a first frame, the model can already see the character, wardrobe, palette, and composition. Do not spend half the prompt describing them again. Tell it what must change after that frame.`}),`
256`,(0,c.jsx)(t.p,{children:`For first-frame generation:`}),`
257`,(0,c.jsxs)(t.blockquote,{children:[`
258`,(0,c.jsx)(t.p,{children:`Preserve the character design and bakery layout. The cake rises another foot and releases flour. The monster catches the spoon, looks at the cake, then smiles. Slow push-in, no new characters.`}),`
259`]}),`
260`,(0,c.jsx)(t.p,{children:`For reference-to-video, state the role of each asset: “use the image for character identity,” “use the audio for beat timing,” or “use the video for camera rhythm.” A reference without a named purpose can pull the generation in several directions at once.`}),`
261`,(0,c.jsx)(t.h2,{children:`Write sound as part of the event`}),`
262`,(0,c.jsx)(t.p,{children:`For models with generated audio, sound belongs beside the action that creates it:`}),`
263`,(0,c.jsxs)(t.blockquote,{children:[`
264`,(0,c.jsx)(t.p,{children:`The tram brakes with a soft metallic squeal as the doors open; low station ambience continues underneath.`}),`
265`]}),`
266`,(0,c.jsx)(t.p,{children:`Keep dialogue short enough to fit the clip. Put exact words in quotation marks and avoid asking one character to deliver a paragraph inside an eight-second shot.`}),`
267`,(0,c.jsx)(t.h2,{children:`Diagnose the result before rewriting everything`}),`
268`,(0,c.jsxs)(t.ul,{children:[`
269`,(0,c.jsxs)(t.li,{children:[(0,c.jsx)(t.strong,{children:`The subject drifts:`}),` add a first frame or reduce simultaneous action.`]}),`
270`,(0,c.jsxs)(t.li,{children:[(0,c.jsx)(t.strong,{children:`The camera wanders:`}),` specify one move and remove competing camera words.`]}),`
271`,(0,c.jsxs)(t.li,{children:[(0,c.jsx)(t.strong,{children:`Nothing happens:`}),` replace mood adjectives with a visible verb and an ending.`]}),`
272`,(0,c.jsxs)(t.li,{children:[(0,c.jsx)(t.strong,{children:`Motion is too fast:`}),` describe a measured pace and reduce the number of beats.`]}),`
273`,(0,c.jsxs)(t.li,{children:[(0,c.jsx)(t.strong,{children:`The scene mutates:`}),` say what must remain unchanged and shorten the transition.`]}),`
274`,(0,c.jsxs)(t.li,{children:[(0,c.jsx)(t.strong,{children:`Audio is crowded:`}),` keep one foreground sound and one ambience bed.`]}),`
275`]}),`
276`,(0,c.jsx)(t.p,{children:`Change one variable per generation. Otherwise, a better result will not tell you which instruction fixed it.`}),`
277`,(0,c.jsx)(t.h2,{children:`A reusable prompt pattern`}),`
278`,(0,c.jsxs)(t.blockquote,{children:[`
279`,(0,c.jsx)(t.p,{children:`[Framing] of [subject] in [place]. [One visible action with an ending]. The camera [one physical move]. [Motion quality and environmental response]. [Lighting and atmosphere]. [Sound if needed]. [Duration and essential constraints].`}),`
280`]}),`
281`,(0,c.jsxs)(t.p,{children:[`This is not a magic formula. It is a checklist that keeps the shot understandable. `,(0,c.jsx)(t.a,{href:`/ai/video`,children:`Try it in the generator`}),`, then save the prompt with the result that moved the way you intended.`]})]})}function D(e={}){let{wrapper:t}={...i(),...e.components};return t?(0,c.jsx)(t,{...e,children:(0,c.jsx)(E,{...e})}):E(e)}var O=e({default:()=>j,meta:()=>k}),k={title:`怎样写出真正“会动”的 AI 视频提示词`,description:`用一套可重复的镜头写法,明确主体动作、摄影机、节奏、光线和声音,而不是把提示词写成一整页剧本。`,created_at:`2026-08-04`,author_name:`Video Lite 编辑部`,author_image:`/logo.svg`,
281image:`/images/showcase/generated/monster-bakery.png`};function A(e){let t={a:`a`,blockquote:`blockquote`,h2:`h2`,li:`li`,p:`p`,strong:`strong`,ul:`ul`,...i(),...e.components};return(0,c.jsxs)(c.Fragment,{children:[(0,c.jsx)(t.p,{children:`图片提示词只需要说明“画面里有什么”,视频提示词还必须说明“接下来发生什么”。如果提示词只有漂亮场景却没有事件,模型就只能自己发明运动,而人物漂移、物体变形和摄影机乱跑,往往就从这里开始。`}),`
282`,(0,c.jsx)(t.p,{children:`解决方法不是写得更长,而是把镜头说得更清楚。`}),`
283`,(0,c.jsx)(t.h2,{children:`用五个决定搭出一个镜头`}),`
284`,(0,c.jsx)(t.p,{children:`建议按这个顺序写:`}),`
285`,(0,c.jsx)(t.p,{children:(0,c.jsx)(t.strong,{children:`主体与地点 → 可见动作 → 摄影机 → 光线与氛围 → 节奏或声音。`})}),`
286`,(0,c.jsx)(t.p,{children:`先看一个只有静态画面的提示词:`}),`
287`,(0,c.jsxs)(t.blockquote,{children:[`
288`,(0,c.jsx)(t.p,{children:`魔法面包房里有一只可爱的橙色小怪物,电影感 3D 动画。`}),`
289`]}),`
290`,(0,c.jsx)(t.p,{children:`它描述了一张图,还没有描述一条视频。把它改成镜头调度:`}),`
291`,(0,c.jsxs)(t.blockquote,{children:[`
292`,(0,c.jsx)(t.p,{children:`一只橙色小怪物站在月光下的石炉旁。圆形蛋糕突然升起,喷出一团柔软的面粉;小怪物后退一步,眨两次眼,然后露出笑容。摄影机从腰部高度缓慢推进。暖色炉火与冷蓝色窗光交织,家庭动画式的轻快节奏,6 秒,无文字。`}),`
293`]}),`
294`,(0,c.jsx)(t.p,{children:`现在每一句都有任务:主体完成动作,摄影机只有一个运动,光线有明确来源,事件也有结尾。`}),`
295`,(0,c.jsx)(t.h2,{children:`主体只安排一个读得懂的动作`}),`
296`,(0,c.jsx)(t.p,{children:`“跳舞、奔跑、转身、大笑、挥手”并不会让镜头更丰富,它只会让五个动作争夺几秒钟。应该选择那个真正改变画面的动作。`}),`
297`,(0,c.jsx)(t.p,{children:`好的动作有清楚的开始和结束:`}),`
298`,(0,c.jsxs)(t.ul,{children:[`
299`,(0,c.jsx)(t.li,{children:`打开雨伞,然后抬头看天;`}),`
300`,(0,c.jsx)(t.li,{children:`跨进列车,车门在身后关闭;`}),`
301`,(0,c.jsx)(t.li,{children:`把玻璃球举到阳光中;`}),`
302`,(0,c.jsx)(t.li,{children:`从窗边转身,与镜头对视;`}),`
303`,(0,c.jsx)(t.li,{children:`从一张纸展开成飞翔的龙。`}),`
304`]}),`
305`,(0,c.jsx)(t.p,{children:`镜头失败时,先减少动作,再考虑更换风格。`}),`
306`,(0,c.jsx)(t.h2,{children:`把摄影机当成真实设备`}),`
307`,(0,c.jsx)(t.p,{children:`能够被真实摄影师执行的摄影机语言,通常更容易得到稳定结果。一次选择一个主要运动:`}),`
308`,(0,c.jsxs)(t.ul,{children:[`
309`,(0,c.jsxs)(t.li,{children:[(0,c.jsx)(t.strong,{children:`固定机位:`}),` 摄影机不动,让主体提供运动。`]}),`
310`,(0,c.jsxs)(t.li,{children:[(0,c.jsx)(t.strong,{children:`推进或拉远:`}),` 改变观众与主体的情绪距离。`]}),`
311`,(0,c.jsxs)(t.li,{children:[(0,c.jsx)(t.strong,{children:`横向跟拍:`}),` 跟随主体穿过空间。`]}),`
312`,(0,c.jsxs)(t.li,{children:[(0,c.jsx)(t.strong,{children:`环绕:`}),` 展示静止主体的形体、尺度或环境。`]}),`
313`,(0,c.jsxs)(t.li,{children:[(0,c.jsx)(t.strong,{children:`手持跟随:`}),` 增加能量和自然的小幅晃动。`]}),`
314`,(0,c.jsxs)(t.li,{children:[(0,c.jsx)(t.strong,{children:`升降或航拍上升:`}),` 在镜头结尾揭示更大的世界。`]}),`
315`]}),`
316`,(0,c.jsx)(t.p,{children:`不要把“无人机高速俯冲、环绕、手持摇晃、缓慢变焦”放在同一个镜头里,这些指令来自不同设备,也意味着不同速度。`}),`
317`,(0,c.jsx)(t.h2,{children:`不只写运动类型,还要写运动质感`}),`
318`,(0,c.jsxs)(t.p,{children:[`模型还需要知道动作应该怎样发生。可以使用 `,(0,c.jsx)(t.strong,{children:`克制、沉重、轻盈、犹豫、被风推动、机械式、几乎不可察觉`}),` 等描述。`]}),`
319`,(0,c.jsx)(t.p,{children:`比较下面两句:`}),`
320`,(0,c.jsxs)(t.blockquote,{children:[`
321`,(0,c.jsx)(t.p,{children:`花朵在风中摇动。`}),`
322`]}),`
323`,(0,c.jsxs)(t.blockquote,{children:[`
324`,(0,c.jsx)(t.p,{children:`高大的罂粟花随着轻微沙漠风,从前景到背景呈不规则的缓慢波浪弯曲,偶尔有花瓣掠过镜头。`}),`
325`]}),`
326`,(0,c.jsx)(t.p,{children:`第二句同时给出了方向、节奏和空间深度。`}),`
327`,(0,c.jsx)(t.h2,{children:`参考素材应该减少描述,而不是增加描述`}),`
328`,(0,c.jsx)(t.p,{children:`上传首帧之后,模型已经看得到角色、服装、配色和构图。不要再用半段提示词重复描述这些内容,直接告诉它这张图之后要发生什么。`}),`
329`,(0,c.jsx)(t.p,{children:`首帧生成可以这样写:`}),`
330`,(0,c.jsxs)(t.blockquote,{children:[`
331`,(0,c.jsx)(t.p,{children:`保持角色设计和面包房布局不变。蛋糕再升高一英尺并喷出面粉,小怪物接住木勺,看向蛋糕,然后微笑。摄影机缓慢推进,不增加新角色。`}),`
332`]}),`
333`,(0,c.jsx)(t.p,{children:`使用参考图片、音频或视频时,要说明每一份素材的职责,例如“图片用于保持角色身份”“音频用于节拍”“视频用于摄影机节奏”。没有职责的参考素材,容易把生成结果同时拉向多个方向。`}),`
334`,(0,c.jsx)(t.h2,{children:`声音要和产生它的事件写在一起`}),`
335`,(0,c.jsx)(t.p,{children:`模型支持生成音频时,声音应该紧跟产生它的动作:`}),`
336`,(0,c.jsxs)(t.blockquote,{children:[`
337`,(0,c.jsx)(t.p,{children:`有轨电车以轻微的金属摩擦声刹停,车门打开,低沉的站台环境声持续铺在下方。`}),`
338`]}),`
339`,(0,c.jsx)(t.p,{children:`对白要短到能够放进视频时长里。把需要准确说出的句子放在引号中,不要让角色在 8 秒镜头里讲一整段台词。`}),`
340`,(0,c.jsx)(t.h2,{children:`重写之前,先判断问题出在哪里`}),`
341`,(0,c.jsxs)(t.ul,{children:[`
342`,(0,c.jsxs)(t.li,{children:[(0,c.jsx)(t.strong,{children:`主体开始漂移:`}),` 加入首帧,或者减少同时发生的动作。`]}),`
343`,(0,c.jsxs)(t.li,{children:[(0,c.jsx)(t.strong,{children:`摄影机乱跑:`}),` 只保留一个摄影机运动,删除相互竞争的镜头词。`]}),`
344`,(0,c.jsxs)(t.li,{children:[(0,c.jsx)(t.strong,{children:`画面什么也没发生:`}),` 减少氛围形容词,增加一个可见动词和结尾。`]}),`
345`,(0,c.jsxs)(t.li,{children:[(0,c.jsx)(t.strong,{children:`动作太快:`}),` 明确写出克制节奏,并减少事件节点。`]}),`
346`,(0,c.jsxs)(t.li,{children:[(0,c.jsx)(t.strong,{children:`场景不断变形:`}),` 说明哪些部分必须保持不变,并缩短转变过程。`]}),`
347`,(0,c.jsxs)(t.li,{children:[(0,c.jsx)(t.strong,{children:`声音太拥挤:`}),` 只保留一个前景声音和一个环境底层。`]}),`
348`]}),`
349`,(0,c.jsx)(t.p,{children:`每次生成只改一个变量。否则即使结果变好,ä¹
349Ÿæ— æ³•判断是哪条指令起了作用。`}),`
350`,(0,c.jsx)(t.h2,{children:`可以重复使用的提示词结构`}),`
351`,(0,c.jsxs)(t.blockquote,{children:[`
352`,(0,c.jsx)(t.p,{children:`[景别]中的[主体]位于[地点]。[一个有结尾的可见动作]。摄影机进行[一个真实可执行的运动]。[动作质感与环境响应]。[光线和氛围]。[必要时写声音]。[时长与关键限制]。`}),`
353`]}),`
354`,(0,c.jsxs)(t.p,{children:[`这不是万能咒语,而是一张保证镜头易于理解的检查表。现在可以`,(0,c.jsx)(t.a,{href:`/ai/video`,children:`在生成器里试一次`}),`,并把真正按预期运动的结果和提示词一起保存下来。`]})]})}function j(e={}){let{wrapper:t}={...i(),...e.components};return t?(0,c.jsx)(t,{...e,children:(0,c.jsx)(A,{...e})}):A(e)}var M=[`choosing-an-ai-video-model`,`how-to-write-ai-video-prompts`,`ai-video-generation-workflows`],N=Object.assign({"/src/content/posts/ai-video-generation-workflows.en.mdx":s,"/src/content/posts/ai-video-generation-workflows.zh.mdx":f,"/src/content/posts/choosing-an-ai-video-model.en.mdx":g,"/src/content/posts/choosing-an-ai-video-model.zh.mdx":b,"/src/content/posts/how-to-write-ai-video-prompts.en.mdx":w,"/src/content/posts/how-to-write-ai-video-prompts.zh.mdx":O});function P(e,t){return M.includes(e)?N[`/src/content/posts/${e}.${t}.mdx`]??N[`/src/content/posts/${e}.en.mdx`]??null:null}function F(e,t){return new Intl.DateTimeFormat(t===`zh`?`zh-CN`:`en-US`,{year:`numeric`,month:t===`zh`?`long`:`short`,day:`numeric`}).format(new Date(e))}var I=r().handler(a(`139369ba08645c08201ff69bd24b596cdc0f7784b8b3ffd8c6a14f47bbeac9cf`)),L=r().handler(a(`b19d1425e0192068330144e802eeba64d1a59e4653703cc76476349d6a195d71`));export{o as a,P as i,I as n,F as r,L as t};

Line numbers count LF bytes from the start of the resource, as the search results do. Vendor segments are library code the classifier recognised; they are stored but not indexed. Bytes are shown as Latin1 characters, one per byte.