Back to Details

video-storyboard

Compare original and translation side by side

🇺🇸

Original

English
🇨🇳

Translation

Chinese

Video Storyboard

视频分镜

Use this skill to create a production-ready storyboard package for a specific scene:
  1. A storyboard image saved to 
    storyboard/scene-XX.png
    .
  2. A video-generation prompt script saved to 
    storyboard/scene-XX.md
    .
Write the skill instructions in English, but create the user-facing 
.md
script in the user's preferred local language unless they explicitly request another language.
使用此Skill为特定场景创建可用于制作的分镜包：
  1. 保存为
    storyboard/scene-XX.png
    的分镜图像。
  2. 保存为
    storyboard/scene-XX.md
    的视频生成提示脚本。
Skill说明使用英文编写，但面向用户的
.md
脚本需使用用户偏好的本地语言，除非用户明确要求其他语言。

Image Generation Requirement

图像生成要求

Storyboard images must be generated by an image-generation model. Do not create storyboard artwork with code, SVG, HTML/CSS, canvas, diagrams, placeholder drawings, or deterministic shape scripts. Those methods are acceptable only for non-visual metadata or file operations, never as a substitute for the storyboard image.
Use the built-in 
image_gen
tool by default, or an approved project image-generation workflow when explicitly required. After generation, save or copy the model-generated raster output to 
storyboard/scene-XX.png
. If the model output cannot be saved directly, locate the generated raster asset and move/copy it into the required path; do not recreate it programmatically.
分镜图像必须由图像生成模型生成。不得使用代码、SVG、HTML/CSS、画布、图表、占位图或确定性形状脚本创建分镜 artwork。这些方法仅适用于非视觉元数据或文件操作，绝不能替代分镜图像。
默认使用内置的
image_gen
工具，或在明确要求时使用已批准的项目图像生成工作流。生成后，将模型生成的栅格输出保存或复制到
storyboard/scene-XX.png
。如果模型输出无法直接保存，找到生成的栅格资源并移动/复制到指定路径；不得通过编程方式重新创建。

When to Use

使用场景

Use this skill when the user asks for any of the following:
  • Generate a storyboard for a scene.
  • Turn a scene from a video plan into storyboard frames.
  • Create prompts for generating video from storyboard panels.
  • Save a storyboard image and prompt script for 
    scene-01
    , 
    scene-02
    , etc.
  • Build a grid storyboard with per-panel timing labels.
If the user only asks for a storyboard image, still consider whether a matching 
.md
prompt script is expected from the workflow. If unclear, ask one concise question or proceed with both files when the user has previously used this workflow.
当用户提出以下任一需求时，使用此Skill：
  • 为某个场景生成分镜。
  • 将视频方案中的场景转换为分镜帧。
  • 创建基于分镜面板生成视频的提示词。
  • 为
    scene-01
    、
    scene-02
    等保存分镜图像和提示脚本。
  • 创建带有单镜头时间标签的网格型分镜。
如果用户仅要求分镜图像，仍需判断工作流是否需要匹配的
.md
提示脚本。若不确定，可提出一个简洁的问题；若用户之前使用过此工作流，则直接生成两个文件。

Inputs to Extract

需要提取的输入信息

Before generating, identify:
  • Scene number.
  • Scene duration.
  • Scene source: existing plan, prior conversation, user prompt, uploaded image, or file.
  • Aspect ratio of the final video, usually 
    16:9
    unless specified.
  • Characters and identity constraints.
  • Available character design assets: 
    .png
    character sheets, 
    .md
    character specs, turnarounds, casting references, or prior approved character images.
  • Wardrobe, props, brand/product details, and location.
  • Visual style and references.
  • Required output language for the 
    .md
    prompt script.
If the scene number is missing, ask for it. If duration is missing, infer it from the referenced plan; otherwise ask or choose a practical default only if the user wants you to proceed.
生成前需确定：
  • 场景编号。
  • 场景时长。
  • 场景来源：已有方案、之前的对话、用户提示、上传的图像或文件。
  • 最终视频的宽高比，未指定时通常为
    16:9
    。
  • 角色及身份约束。
  • 可用的角色设计资源：
    .png
    角色表、
    .md
    角色规格、多角度视图、选角参考或已批准的过往角色图像。
  • 服装、道具、品牌/产品细节及场景位置。
  • 视觉风格及参考示例。
  • .md
    提示脚本的输出语言要求。
若缺少场景编号，需询问用户。若缺少时长，可从参考方案中推断；否则需询问用户，或仅在用户要求继续时选择合理默认值。

Output Paths

输出路径

Always use a two-digit scene number:
  • Scene 1 -> 
    storyboard/scene-01.png
    and 
    storyboard/scene-01.md
  • Scene 2 -> 
    storyboard/scene-02.png
    and 
    storyboard/scene-02.md
  • Scene 12 -> 
    storyboard/scene-12.png
    and 
    storyboard/scene-12.md
Create 
storyboard/
if it does not exist.
Do not overwrite existing storyboard files unless the user asked for regeneration or replacement. If a target file exists, either ask before replacing or write a clear revision path such as 
scene-01-v2.png
if the user requested alternatives.
始终使用两位数字的场景编号：
  • 场景1 -> 
    storyboard/scene-01.png
    和 
    storyboard/scene-01.md
  • 场景2 -> 
    storyboard/scene-02.png
    和 
    storyboard/scene-02.md
  • 场景12 -> 
    storyboard/scene-12.png
    和 
    storyboard/scene-12.md
若
storyboard/
目录不存在则创建。
除非用户要求重新生成或替换，否则不得覆盖现有分镜文件。若目标文件已存在，替换前需询问用户；若用户要求提供替代版本，可使用清晰的修订路径，如
scene-01-v2.png
。

Grid Rules

网格规则

Choose grid size from the scene duration:
Scene DurationGridPanel Count
<= 9s
3 rows x 3 columns
9 panels
> 9s
and 
<= 12s
4 rows x 3 columns
12 panels
> 12s
4 rows x 4 columns
16 panels
Each panel must be 
16:9
.
There must be no gaps between panels. Use thin black panel borders only if needed for readability, but do not insert gutters or spacing.
If the duration does not divide evenly, allocate practical panel durations. Prefer simple labels such as 
0.5s
, 
1s
, 
1.5s
, 
2s
, or exact ranges when helpful. Empty final panels are allowed for very short scenes, but only when filling every panel would over-fragment the action. Empty panels should be plain black.
根据场景时长选择网格尺寸：
场景时长网格布局镜头数量
<= 9s
3行×3列
9个镜头
> 9s
且 
<= 12s
4行×3列
12个镜头
> 12s
4行×4列
16个镜头
每个镜头的宽高比必须为
16:9
。
镜头之间不得有间隙。仅在需要提升可读性时使用细黑边框，不得添加 gutter（ gutter 保留原术语）或间距。
若时长无法平均分配，需设置合理的镜头时长。优先使用简单标签，如
0.5s
、
1s
、
1.5s
、
2s
，必要时使用精确范围。极短场景允许最后几个镜头为空，但仅在填满所有镜头会过度拆分动作时使用。空镜头应为纯黑色。

Storyboard Image Requirements

分镜图像要求

Generate one single storyboard sheet, not separate images.
The storyboard sheet itself must come from the image-generation model. The prompt should explicitly request the complete grid, panel labels, continuity constraints, and visual style. Do not implement the grid as SVG or draw panels manually with code.
Before generating any storyboard image, first look for existing character images in the current conversation/context, including images that were generated earlier in the project. Then search the current project for matching character assets under 
characters/
or 
storyboard/
, especially 
.png
files, character sheets, turnarounds, and approved reference images.
If any relevant character image exists, it must be passed to the image-generation model as a reference image for the storyboard. Do not rely on text description alone when a character image is available.
If the current context includes a character design image, approved character sheet, turnaround sheet, or character reference 
.png
, use it as an image reference for the storyboard image generation. This is required to preserve identity, face, hairstyle, body proportions, wardrobe, and material details across scenes.
If both a character 
.md
spec and 
.png
sheet exist, read the 
.md
for written constraints and feed the 
.png
to the image-generation model as the visual reference.
Every visible panel must include:
  • A black rectangle in the top-left corner with the panel number in white text, starting from 
    1
    .
  • A black rectangle in the top-right corner with the panel duration in white text, such as 
    0.5s
    , 
    1s
    , or 
    1.5s
    .
  • A clear shot composition that corresponds to the panel's beat.
Panel variety is critical:
  • Avoid creating two panels that are identical, near-identical, or visually redundant.
  • Each non-empty panel should advance the action, reveal new information, or change the viewer's perspective.
  • Vary shot size, camera angle, character pose, gesture, prop interaction, foreground/background relationship, or motion direction while preserving continuity.
  • Do not use repeated close-ups or repeated wide shots unless the repetition is intentional and clearly serves the scene rhythm.
Critical continuity requirements:
  • Keep the same character identity across all panels.
  • Keep face, age, hairstyle, body type, and skin tone consistent.
  • Keep wardrobe consistent unless the scene explicitly includes a costume change.
  • Keep key props consistent, including product shape, color, bottle/can design, bag, camera, vehicle, or luggage.
  • Keep location geography consistent: doors, windows, roads, vending machines, classrooms, landmarks, field direction, and light direction should not jump randomly.
  • Keep the time of day and weather consistent unless the scene explicitly transitions.
When generating the image prompt, state continuity requirements explicitly and repeatedly enough for the image model to respect them. This matters more than decorative styling.
When using a character design reference image, explicitly tell the image model that the storyboard character must match that reference: same face, hairstyle, age, body proportions, outfit, shoes, accessories, clothing material, and color palette. The scene may change pose, action, camera angle, and lighting, but not the character design.
Avoid:
  • AI-looking faces, waxy skin, distorted hands, extra fingers.
  • Random changes in clothing, hair, props, or background.
  • Uncontrolled readable text or broken brand typography.
  • Extra logos, subtitles, speech bubbles, UI overlays, or watermarks.
  • Overly cinematic fantasy lighting if the scene should feel real.
For brand/product work, prefer "brand-color and product-form cues" unless the user provides official assets. Generated text on packaging is often unreliable.
生成单个分镜页，而非单独图像。
分镜页本身必须由图像生成模型生成。提示词需明确要求完整的网格、镜头标签、连贯性约束及视觉风格。不得使用SVG实现网格或通过代码手动绘制镜头。
生成任何分镜图像前，先在当前对话/上下文查找现有角色图像，包括项目中早期生成的图像。然后在当前项目的
characters/
或
storyboard/
目录下搜索匹配的角色资源，尤其是
.png
文件、角色表、多角度视图和已批准的参考图像。
若存在相关角色图像，必须将其作为参考图像传递给图像生成模型。当有角色图像可用时，不得仅依赖文字描述。
若当前上下文包含角色设计图像、已批准的角色表、多角度视图或角色参考
.png
，需将其作为分镜图像生成的视觉参考。这是确保跨场景角色身份、面部、发型、身材比例、服装及材质细节一致性的必要要求。
若同时存在角色
.md
规格和
.png
表，需读取
.md
中的书面约束，并将
.png
作为视觉参考提供给图像生成模型。
每个可见镜头必须包含：
  • 左上角的黑色矩形框，内有白色镜头编号，从
    1
    开始。
  • 右上角的黑色矩形框，内有白色镜头时长，如
    0.5s
    、
    1s
    或
    1.5s
    。
  • 与镜头节拍对应的清晰镜头构图。
镜头多样性至关重要：
  • 避免创建相同、近乎相同或视觉冗余的两个镜头。
  • 每个非空镜头都应推进动作、揭示新信息或改变观众视角。
  • 在保持连贯性的同时，改变镜头大小、拍摄角度、角色姿势、手势、道具互动、前景/背景关系或运动方向。
  • 除非重复是有意且明确服务于场景节奏，否则不得使用重复的特写或重复的广角镜头。
关键连贯性要求：
  • 所有镜头中的角色身份保持一致。
  • 面部、年龄、发型、体型和肤色保持一致。
  • 服装保持一致，除非场景明确包含服装变化。
  • 关键道具保持一致，包括产品形状、颜色、瓶/罐设计、包、相机、车辆或行李。
  • 场景地理位置保持一致：门、窗、道路、自动售货机、教室、地标、场地方向和光线方向不得随意变化。
  • 时间和天气保持一致，除非场景明确包含过渡。
生成图像提示词时，需明确且反复强调连贯性要求，确保图像模型遵循。这比装饰性风格更重要。
使用角色设计参考图像时，需明确告知图像模型：分镜中的角色必须与参考图像匹配，包括相同的面部、发型、年龄、身材比例、服装、鞋子、配饰、服装材质和调色板。场景可改变姿势、动作、拍摄角度和光线，但不得改变角色设计。
需避免：
  • AI感面部、蜡状皮肤、扭曲的手、多余手指。
  • 服装、发型、道具或背景的随机变化。
  • 不受控的可读文本或破损的品牌字体。
  • 额外的标志、字幕、对话气泡、UI叠加层或水印。
  • 若场景需真实感，则避免过于戏剧化的奇幻光线。
对于品牌/产品相关工作，优先使用“品牌色彩和产品形态提示”，除非用户提供官方资源。包装上生成的文本通常不可靠。

Recommended Image Prompt Structure

推荐的图像提示词结构

Use this structure for the image generation prompt:
text
Create a single storyboard sheet for Scene [XX], [duration] seconds, [video aspect/style].

Use the provided character design image as the primary visual reference for the character. Match the same face, hairstyle, body proportions, wardrobe, shoes, accessories, clothing material details, and color palette across all panels.

Grid: [3x3 / 4x3 / 4x4], every panel is 16:9, no gaps between panels.
Panel labels: top-left black label with white panel number; top-right black label with white duration.

Continuity is critical: the same [character description] appears consistently across all panels, same face, same hairstyle, same wardrobe, same props, same location geography, same lighting direction, same time of day.

Panel variety is critical: no two non-empty panels should look identical or overly similar. Each panel must have a distinct shot composition, action beat, camera angle, shot size, gesture, prop interaction, or foreground/background arrangement while still preserving continuity.

Scene summary: [one concise scene summary].
Visual style: [photorealistic / cinematic / documentary / commercial / handheld / natural light].

Panels:
1. [duration] - [shot type, action, composition]
2. [duration] - [shot type, action, composition]
...

Negative constraints: no subtitles, no extra logos, no watermarks, no distorted text, no inconsistent character, no clothing changes, no prop changes, no AI-smooth skin, no deformed hands.
使用以下结构编写图像生成提示词：
text
Create a single storyboard sheet for Scene [XX], [duration] seconds, [video aspect/style].

Use the provided character design image as the primary visual reference for the character. Match the same face, hairstyle, body proportions, wardrobe, shoes, accessories, clothing material details, and color palette across all panels.

Grid: [3x3 / 4x3 / 4x4], every panel is 16:9, no gaps between panels.
Panel labels: top-left black label with white panel number; top-right black label with white duration.

Continuity is critical: the same [character description] appears consistently across all panels, same face, same hairstyle, same wardrobe, same props, same location geography, same lighting direction, same time of day.

Panel variety is critical: no two non-empty panels should look identical or overly similar. Each panel must have a distinct shot composition, action beat, camera angle, shot size, gesture, prop interaction, or foreground/background arrangement while still preserving continuity.

Scene summary: [one concise scene summary].
Visual style: [photorealistic / cinematic / documentary / commercial / handheld / natural light].

Panels:
1. [duration] - [shot type, action, composition]
2. [duration] - [shot type, action, composition]
...

Negative constraints: no subtitles, no extra logos, no watermarks, no distorted text, no inconsistent character, no clothing changes, no prop changes, no AI-smooth skin, no deformed hands.

Video Prompt Script Requirements

视频提示脚本要求

After creating the image, write a matching 
.md
file for video generation.
The script should let a video model generate a single continuous scene from the storyboard. It should describe continuity, style, shot timing, character action, sound design, editing rhythm, and negative prompts.
Use the user's preferred local language for the script. If the user wrote in Chinese, write the 
.md
in Chinese.
Use this structure:
markdown
根据上传的 01-[NN] 分镜图，生成一段 [duration] 秒 [style] 场景。

[Continuity paragraph: same characters, wardrobe, props, location, time of day.]

[Restrictions paragraph: no subtitles, no voiceover unless requested, no extra logos, no readable text if risky.]

风格：[visual style, camera feel, color, texture.]

节奏：[overall edit rhythm.]

---
创建图像后，编写匹配的
.md
文件用于视频生成。
脚本应能让视频模型根据分镜生成单个连续场景。需描述连贯性、风格、镜头时间、角色动作、声音设计、剪辑节奏和负向提示词。
脚本使用用户偏好的本地语言。若用户使用中文提问，则
.md
脚本使用中文。
使用以下结构：
markdown
根据上传的 01-[NN] 分镜图，生成一段 [duration] 秒 [style] 场景。

[连贯性段落：相同角色、服装、道具、场景、时间。]

[限制段落：无字幕，除非要求否则无旁白，无额外标志，若有风险则无可读文本。]

风格：[视觉风格、拍摄质感、色彩、纹理。]

节奏：[整体剪辑节奏。]

---

01 | 00:00-00:00.5

01 | 00:00-00:00.5

[Panel description.]
[Character action.]
[Camera movement / shot type / transition.]
[镜头描述。]
[角色动作。]
[镜头运动 / 镜头类型 / 转场方式。]

02 | 00:00.5-00:01.0

02 | 00:00.5-00:01.0

[...]
[...]

声音设计

声音设计

[Natural sound, music, transitions, sound bridges.]
[自然音效、音乐、转场音效、声音衔接。]

剪辑建议

剪辑建议

[How the panels should connect into motion.]
[镜头如何衔接成动态画面。]

负向提示词

负向提示词

  • [No subtitles]
  • [No extra logos]
  • [No inconsistent character]
  • [No wardrobe changes]
  • [No prop changes]
  • [No unreadable or distorted text]

Translate the headings if the user requested English or another language.
  • [无字幕]
  • [无额外标志]
  • [无角色不一致]
  • [无服装变化]
  • [无道具变化]
  • [无不可读或扭曲文本]

若用户要求英文或其他语言，需翻译标题。

Timing Rules for the Script

脚本时间规则

The 
.md
script must match the storyboard image:
  • Same panel count as the grid, unless some final panels are intentionally blank.
  • Same per-panel durations as the image labels.
  • Timecodes add up exactly to the scene duration.
  • Each panel describes a single visual beat, not a full mini-scene.
  • Fast scenes can use 0.5s beats; slower scenes can use 1s-2s beats.
For a 5-second 3x3 board, a good default is:
  • Panels 1-8: 
    0.5s
    each.
  • Panel 9: 
    1s
    .
For a 9-second 3x3 board, a good default is:
  • Panels 1-9: 
    1s
    each.
For a 12-second 4x3 board, a good default is:
  • Panels 1-12: 
    1s
    each.
For a 15-second 4x4 board, a good default is:
  • Panels 1-14: 
    1s
    each.
  • Panel 15: 
    0.5s
    .
  • Panel 16: 
    0.5s
    .
Adjust these when the scene's action needs different emphasis.
.md
脚本必须与分镜图像匹配：
  • 网格镜头数量相同，除非最后几个镜头是故意留白。
  • 每个镜头的时长与图像标签一致。
  • 时间码总和恰好等于场景时长。
  • 每个镜头描述单个视觉节拍，而非完整的小场景。
  • 快节奏场景可使用0.5秒节拍；慢节奏场景可使用1-2秒节拍。
对于5秒的3×3分镜，合理默认设置为：
  • 镜头1-8：每个
    0.5s
    。
  • 镜头9：
    1s
    。
对于9秒的3×3分镜，合理默认设置为：
  • 镜头1-9：每个
    1s
    。
对于12秒的4×3分镜，合理默认设置为：
  • 镜头1-12：每个
    1s
    。
对于15秒的4×4分镜，合理默认设置为：
  • 镜头1-14：每个
    1s
    。
  • 镜头15：
    0.5s
    。
  • 镜头16：
    0.5s
    。
可根据场景动作的重点调整上述设置。

Workflow

工作流

  1. Read the relevant plan, previous scene script, existing storyboard, or user-provided scene description.
  2. Search the current conversation/context for existing character images, including images generated earlier in the project or attached by the user.
  3. Search the current project's 
    characters/
    folder for corresponding character images, especially 
    characters/*.png
    , approved character sheets, turnarounds, and matching 
    {character-name}.md
    specs.
  4. If a relevant character design image exists in either place, include it as a reference image when generating the storyboard. This is required for character consistency.
  5. Extract the scene's duration, story purpose, characters, wardrobe, props, location, camera language, and edit beat.
  6. Choose the grid size from the duration.
  7. Draft panel beats with exact durations and timecodes.
  8. Generate the storyboard image with the image-generation model/tool, using any available character design image as a visual reference. Do not replace this with SVG, code drawing, canvas, HTML/CSS, or placeholder art.
  9. Save or copy the generated image to 
    storyboard/scene-XX.png
    .
  10. Inspect the image if possible. Verify grid, labels, continuity, and obvious visual artifacts.
  11. Write 
    storyboard/scene-XX.md
    using the same panel timing and descriptions.
  12. Final response should include the saved file paths and, when supported, render the image with a Markdown image tag using an absolute path.
  1. 读取相关方案、过往场景脚本、现有分镜或用户提供的场景描述。
  2. 在当前对话/上下文查找现有角色图像，包括项目中早期生成的图像或用户上传的图像。
  3. 在当前项目的
    characters/
    文件夹中搜索对应的角色图像，尤其是
    characters/*.png
    、已批准的角色表、多角度视图和匹配的
    {character-name}.md
    规格。
  4. 若上述任一位置存在相关角色设计图像，生成分镜时需将其作为参考图像。这是确保角色一致性的必要步骤。
  5. 提取场景的时长、叙事目的、角色、服装、道具、场景位置、镜头语言和剪辑节拍。
  6. 根据时长选择网格尺寸。
  7. 草拟带有精确时长和时间码的镜头节拍。
  8. 使用图像生成模型/工具生成分镜图像，使用任何可用的角色设计图像作为视觉参考。不得使用SVG、代码绘图、画布、HTML/CSS或占位图替代。
  9. 将生成的图像保存或复制到
    storyboard/scene-XX.png
    。
  10. 尽可能检查图像，验证网格、标签、连贯性和明显的视觉瑕疵。
  11. 使用相同的镜头时间和描述编写
    storyboard/scene-XX.md
    。
  12. 最终回复应包含保存的文件路径，若支持，使用Markdown图片标签和绝对路径渲染图像。

Quality Checklist

质量检查清单

Before finishing, verify:
  • storyboard/scene-XX.png
    exists.
  • storyboard/scene-XX.md
    exists.
  • The scene number is two digits.
  • The grid size matches the duration rule.
  • The current context was checked for existing/generated character images.
  • The project 
    characters/
    folder was checked for corresponding character images.
  • Any available character design image from context or 
    characters/
    was used as a reference image for generation.
  • Every non-empty panel has a number label and duration label.
  • No two non-empty panels are identical, near-identical, or visually redundant.
  • The panel durations add up to the scene duration.
  • The 
    .md
    script matches the image panel order.
  • Character continuity is explicitly stated in both the image prompt and 
    .md
    script.
  • Wardrobe, props, and location consistency are explicitly stated.
  • The output language of the 
    .md
    script matches the user's preference.
完成前需验证：
  • storyboard/scene-XX.png
    已存在。
  • storyboard/scene-XX.md
    已存在。
  • 场景编号为两位数字。
  • 网格尺寸符合时长规则。
  • 已检查当前上下文是否存在现有/生成的角色图像。
  • 已检查项目
    characters/
    文件夹是否存在对应的角色图像。
  • 上下文或
    characters/
    中可用的角色设计图像已作为生成参考图像。
  • 每个非空镜头都有编号标签和时长标签。
  • 没有两个非空镜头是相同、近乎相同或视觉冗余的。
  • 镜头时长总和等于场景时长。
  • .md
    脚本与图像镜头顺序匹配。
  • 图像提示词和
    .md
    脚本中均明确说明角色连贯性。
  • 明确说明服装、道具和场景位置的一致性。
  • .md
    脚本的输出语言符合用户偏好。

Example: 5-Second Commercial Scene

示例：5秒广告场景

User asks:
text
Create storyboard and video prompt for Scene 01: 5-second campus sports drink TVC opener.
Use:
  • Grid: 
    3x3
    .
  • Panels: 9.
  • Timing: first 8 panels 
    0.5s
    , final panel 
    1s
    .
  • Output files: 
    storyboard/scene-01.png
    , 
    storyboard/scene-01.md
    .
  • Continuity emphasis: same student, same school uniform, same blue-white drink bottle, same morning campus location.
Example panel plan:
PanelDurationBeat
10.5sVending machine close-up in morning campus light
20.5sBottle drops into pickup slot
30.5sStudent's hand reaches for bottle
40.5sBottle lifted, condensation catches light
50.5sStudent looks toward sports field
60.5sWide shot of school gate
70.5sStudent starts moving along campus path
80.5sLow-angle sneaker first running step
91sSuper-wide reveal toward track field
用户提问：
text
Create storyboard and video prompt for Scene 01: 5-second campus sports drink TVC opener.
使用：
  • 网格：
    3×3
    。
  • 镜头：9个。
  • 时间：前8个镜头每个
    0.5s
    ，最后一个镜头
    1s
    。
  • 输出文件：
    storyboard/scene-01.png
    、
    storyboard/scene-01.md
    。
  • 连贯性重点：同一学生、同一校服、同一蓝白配色饮料瓶、同一清晨校园场景。
示例镜头规划：
镜头时长节拍
10.5s清晨校园光线中的自动售货机特写
20.5s饮料瓶掉落至取货口
30.5s学生伸手去拿饮料瓶
40.5s拿起饮料瓶，冷凝反光
50.5s学生望向运动场
60.5s校门广角镜头
70.5s学生开始沿校园小路行走
80.5s低角度拍摄运动鞋迈出第一步
91s超广角镜头展现跑道