Skip to content
Back to search

gc-still-image-motion-director

Analyze still images and design restrained, image-specific motion for image-to-video generation. Use when the user provides one or more images, screenshots, posters, collages, illustrations, editorial graphics, or an image folder and asks how the image should move, which elements must stay fixed, whether it should move at all, or how to write a prompt for Jimeng or another image-to-video model. Also use for requests about preventing text, paper, layouts, frames, or graphic elements from drifting during animation. Do not use for ordinary image generation, image editing, or analysis of an existing finished video unless the user wants a reusable still-image motion prompt from it.

Safety scan

A static text scan, not a guarantee. Select the badge to see what was scanned and every finding.

Checked today
Badge for your README
findsafeskills safety rating

Markdown

[![findsafeskills safety rating](/api/listings/liamgvchi-gc-still-image-motion-director/badge.svg)](/listings/liamgvchi-gc-still-image-motion-director)

HTML

<a href="/listings/liamgvchi-gc-still-image-motion-director"><img src="/api/listings/liamgvchi-gc-still-image-motion-director/badge.svg" alt="findsafeskills safety rating" /></a>

Install

i

Install instructions from the README

As written by the repo's author. findsafeskills has not verified these steps.

Clone the public repository directly into the Codex skills directory:

git clone https://github.com/LiamGvchi/gc-still-image-motion-director.git \
  ~/.codex/skills/gc-still-image-motion-director

Restart Codex if the skill does not appear immediately.

What was scanned

The files the safety scan read, as written by the repository's author and shown as plain text. Lines with a finding are marked.

SKILL.md · Manifest · 120 lines

View this file on GitHub (read at the repository's current version: the scan did not record which commit it read)

  1. ---
  2. name: gc-still-image-motion-director
  3. description: Analyze still images and design restrained, image-specific motion for image-to-video generation. Use when the user provides one or more images, screenshots, posters, collages, illustrations, editorial graphics, or an image folder and asks how the image should move, which elements must stay fixed, whether it should move at all, or how to write a prompt for Jimeng or another image-to-video model. Also use for requests about preventing text, paper, layouts, frames, or graphic elements from drifting during animation. Do not use for ordinary image generation, image editing, or analysis of an existing finished video unless the user wants a reusable still-image motion prompt from it.
  4. ---
  5. # GC Still Image Motion Director
  6. Turn a still image into a motion decision and a copy-ready image-to-video prompt. Derive motion from the image's subject, meaning, composition, and physical relationships; do not apply a stock animation effect.
  7. ## Core Workflow
  8. 1. Inspect the real image before proposing motion.
  9. - For a folder, inventory readable image files, dimensions, names, and visual patterns. Create a contact sheet when it materially improves comparison.
  10. - If an image cannot be accessed, ask the user to attach or provide it again instead of inventing its contents.
  11. 2. Separate observation from recommendation.
  12. - State what is visibly present.
  13. - Label inferred meaning or motion intent as a recommendation, not a fact.
  14. 3. Identify:
  15. - primary subject;
  16. - subject type and implied behavior;
  17. - composition axis, focal point, whitespace, frames, and hierarchy;
  18. - text, numbers, paper texture, annotations, borders, and other fragile layout elements;
  19. - physical relationships such as gravity, tension, wind, water, contact, rotation, or resistance.
  20. 4. Decide whether the image should use:
  21. - `motion`: one clear restrained action;
  22. - `micro-motion`: an almost still action or environmental trace;
  23. - `static`: no subject motion; explain why stillness serves the image better.
  24. 5. Select at most:
  25. - one primary motion;
  26. - one physically related secondary motion;
  27. - one optional observation effect, only when justified.
  28. 6. Define the lock list before writing the prompt.
  29. 7. Write a platform-ready prompt with a concrete subject, path, range, timing, locks, and failure prohibitions.
  30. 8. Check that the proposal preserves the original design rather than regenerating it frame by frame.
  31. Read [references/motion-decision-framework.md](references/motion-decision-framework.md) when the image's correct motion is ambiguous, when comparing several images, or when deciding that a subject should remain still.
  32. Read [references/prompt-construction.md](references/prompt-construction.md) when writing the final prompt, adapting it to a named platform, or diagnosing why a previous prompt caused drift or deformation.
  33. ## Motion Rule
  34. Let motion come from the image itself.
  35. Prefer, in order:
  36. 1. an intrinsic action of the subject;
  37. 2. a physical response caused by that action;
  38. 3. an environmental or observation effect when the first two are unsuitable.
  39. Do not default to scanning light, floating, breathing paper, parallax, particles, camera push-in, or generic cinematic motion. Do not force a specimen, stone, archive, paused mechanism, or quiet object to behave like a living subject. A valid result may recommend no subject movement.
  40. Keep motion restrained unless the user explicitly asks for stronger animation. Use the smallest movement that makes the intended relationship legible.
  41. ## Stability Rules
  42. - Lock typography, numbers, labels, handwritten notes, paper texture, borders, frames, grids, and collage geometry.
  43. - Keep subjects inside their original visual region unless the composition clearly provides a path.
  44. - Preserve subject count, identity, silhouette, clothing, proportions, and object connectivity.
  45. - Keep ropes, handles, joints, stems, limbs, and mechanical links continuous.
  46. - Avoid camera movement by default for poster, collage, archive, and editorial layouts.
  47. - Do not describe unrelated styling that is already supplied by the source image; redundant style language can encourage redraw.
  48. - For a motionless recommendation, propose a static export or a minimal environmental trace instead of inventing action.
  49. ## Output
  50. For one image, return:
  51. ```markdown
  52. 图片:
  53. [name]
  54. 画面观察:
  55. [visible subject, composition, and fragile elements]
  56. 动态判断:
  57. [motion / micro-motion / static] — [short reason]
  58. 建议运动:
  59. - 主动作:
  60. - 伴随动作:
  61. - 范围与路径:
  62. - 节奏与时长:
  63. 必须固定:
  64. - [...]
  65. 不建议:
  66. - [...]
  67. 可直接复制到[目标平台]:
  68. [complete prompt]
  69. 主要风险:
  70. - [...]
  71. ```
  72. Omit empty fields. If the user asks for only a prompt, return only the copy-ready prompt.
  73. For a folder or image set:
  74. 1. summarize the shared visual system;
  75. 2. state which constraints can be shared;
  76. 3. classify each image as `motion`, `micro-motion`, or `static`;
  77. 4. write a separate prompt for each requested image;
  78. 5. do not claim that one universal prompt is sufficient when subject behavior differs.
  79. Default to Chinese when the user writes Chinese. Default to a restrained 4-second image-to-video clip when the target model and duration are not specified, and state that assumption briefly.
  80. ## Quality Check
  81. Before delivering, verify:
  82. - Is the motion caused by something visible in the image?
  83. - Is there only one dominant action?
  84. - Does the secondary motion obey the same physical cause?
  85. - Are all fragile design elements explicitly locked?
  86. - Does the prompt specify range, direction, and rhythm rather than saying only "move naturally"?
  87. - Could the result still work if the camera never moves?
  88. - Did you avoid making every image use the same effect?
  89. - Did you mark uncertainty instead of presenting a guessed original prompt as fact?
README.md · README · 155 lines

View this file on GitHub (read at the repository's current version: the scan did not record which commit it read)

  1. # GC Still Image Motion Director
  2. **English** · [简体中文](README.zh-CN.md)
  3. A Codex skill that analyzes a real still image, decides whether it should use motion, micro-motion, or remain static, and writes a restrained image-to-video prompt without breaking typography, paper texture, framing, or layout.
  4. The callable skill name is `gc-still-image-motion-director`.
  5. This skill writes motion direction and a copy-ready prompt. It does not generate the final video.
  6. ## What It Does
  7. The skill:
  8. - reads the actual image before proposing motion;
  9. - separates visible observations from recommendations;
  10. - selects at most one primary action and one physically related response;
  11. - explicitly locks text, numbers, paper, frames, grids, and composition;
  12. - supports `motion`, `micro-motion`, and `static` decisions;
  13. - adapts the final prompt to a named image-to-video platform when requested;
  14. - diagnoses text, paper, camera, and layout drift in failed results.
  15. It avoids stock effects such as automatic camera push-ins, floating layers, breathing paper, generic particles, and unsupported parallax.
  16. ## Examples
  17. These two supplied cases show the intended difference between `motion` and `micro-motion`.
  18. For each case, the user supplied the source image and asked Codex to use this Skill. Codex inspected the image and generated the Chinese Seedance 2.0 prompt shown below; that prompt was then used with Seedance 2.0 to generate the linked result clip. The clips are real outputs from those runs. Generative results can vary, and a returned take may not obey every lock perfectly.
  19. ### `motion` — CHOICE
  20. <img src="examples/choice-butterfly.jpg" width="360" alt="CHOICE editorial poster with a butterfly beside a blue rectangle">
  21. The butterfly stays in its original focal region and completes one bounded wingbeat. The blue rectangle, envelope drawing, typography, Chinese copy, paper texture, and full composition remain fixed.
  22. [Watch the 4-second MP4 result](examples/choice-butterfly.mp4)
  23. The source was a verified Apple Live Photo pair. This repository uses a silent MP4 derivative for browser compatibility and does not publish the Live Photo package.
  24. <details>
  25. <summary>Seedance 2.0 prompt used for this result (Chinese)</summary>
  26. ```text
  27. 固定镜头,保持原图构图、纸张质感和全部排版完全不变。
  28. 只让蓝色矩形左上方现有的白色线稿蝴蝶缓慢完成一次振翅:双翼从当前展开状态向内收拢约8至10度,再平稳展开回到原始角度。蝴蝶身体和落点基本不动,整体位移不超过蓝色矩形宽度的1%,最后完全停稳。
  29. 蓝色矩形、信封线稿、信封折线、中央火漆印、三个浅蓝圆点、CHOICE及其倒影、中文“命运偶尔借一只蝴蝶,改写人的选择。”、四周日期、编号、档案文字、边线、折痕、污点和纸张纹理全部固定。
  30. 禁止蝴蝶飞走、持续扇翅、增加或减少翅膀,禁止蝴蝶复制或变形成真实昆虫,禁止信封打开、移动或变形,禁止火漆印旋转,禁止圆点漂浮,禁止文字变化、纸张呼吸和镜头移动。
  31. 4秒,先停留约0.6秒,只完成一次缓慢振翅,最后恢复静止。
  32. ```
  33. </details>
  34. ### `micro-motion` — Walk Slowly
  35. <img src="examples/walk-slowly.png" width="360" alt="Minimal editorial poster with grass and cloud photo windows">
  36. Motion remains inside the two photo windows: light and grass shift subtly in the left image while the cloud changes gently in the right image. The path text, Chinese title, paper texture, spacing, and full layout remain fixed.
  37. [Watch the 4-second MP4 result](examples/walk-slowly.mp4)
  38. <details>
  39. <summary>Seedance 2.0 prompt used for this result (Chinese)</summary>
  40. ```text
  41. 固定镜头,完整保持原图的纵向海报构图、留白和编辑排版。
  42. 前1秒画面保持静止。随后只让右侧云朵照片框内部的白云缓慢向右漂移,位移不超过该照片框宽度的3%,云朵形状、体积和边缘保持稳定,不生成新的云朵。
  43. 同时,左侧草地照片框内部受到同一阵轻微微风影响,少量草叶只摆动2–3度,斑驳阳光在草地表面产生非常缓慢、细微的明暗变化。所有动态必须严格限制在两个矩形照片框内部,最后1秒逐渐稳定下来。
  44. 绿色曲线、绿色圆点、“walk slowly today with the wind”全部单词、右侧浅灰英文诗句、中文标题、日期、署名、照片框边界、旧纸颜色、纸张纤维、颗粒、污点和大面积留白完全固定。
  45. 禁止文字改写、字母闪烁、绿色曲线变形、圆点移动、照片框漂移或变形;禁止云朵快速移动、翻滚或新增;禁止草地大幅摇晃;禁止纸张呼吸、纹理漂浮、全画面曝光变化、镜头推近、平移、缩放、旋转、景深变化和视差。
  46. 4秒,极度克制、安静,像一阵微风只经过照片里的草地和云,海报本身始终保持静止。
  47. ```
  48. </details>
  49. ## Installation
  50. Clone the public repository directly into the Codex skills directory:
  51. ```bash
  52. git clone https://github.com/LiamGvchi/gc-still-image-motion-director.git \
  53. ~/.codex/skills/gc-still-image-motion-director
  54. ```
  55. Restart Codex if the skill does not appear immediately.
  56. ## Usage
  57. Invoke the skill with a still image:
  58. ```text
  59. 用 $gc-still-image-motion-director 分析这张海报应该怎么动,
  60. 锁住文字和纸张,并给我一条可复制的图生视频 Prompt。
  61. ```
  62. For a folder of images:
  63. ```text
  64. 用 $gc-still-image-motion-director 分析这个图片文件夹,
  65. 分别判断哪些适合动、哪些只适合微动、哪些应该保持静止。
  66. ```
  67. To repair a failed result:
  68. ```text
  69. 用 $gc-still-image-motion-director 重写这条 Prompt:
  70. 上一版出现了文字变化、纸张呼吸和镜头漂移。
  71. ```
  72. ## Output
  73. For each image, the skill can return:
  74. 1. visible image observations;
  75. 2. a `motion`, `micro-motion`, or `static` decision;
  76. 3. one bounded primary action and an optional linked response;
  77. 4. an explicit lock list;
  78. 5. a copy-ready image-to-video prompt;
  79. 6. the main predicted failure risks.
  80. The default is a restrained 4-second clip when no platform or duration is specified.
  81. ## Repository Structure
  82. - `SKILL.md`: the core workflow and output contract
  83. - `references/motion-decision-framework.md`: motion suitability, amplitude, timing, and composition rules
  84. - `references/prompt-construction.md`: prompt order, lock lists, negative constraints, and repair guidance
  85. - `agents/openai.yaml`: Codex UI metadata
  86. - `evals/evals.json`: representative trigger and non-trigger cases
  87. - `examples/`: two supplied still-and-result pairs demonstrating `motion` and `micro-motion`
  88. - `LICENSE`: MIT license
  89. This repository publishes one standalone skill. It does not include private backups, unrelated user media, platform credentials, or claims that the Skill itself generated the example videos.
  90. ## Compatibility Notice
  91. This project is independent and is not affiliated with or endorsed by Jimeng or any image-to-video platform. Product names are used only to describe compatibility.
  92. Platform behavior may change. The skill does not claim undocumented controls and falls back to a platform-neutral prompt when behavior is uncertain.
  93. ## Related Skill
  94. For quiet minimal zine-style poster generation, see [GC Minimal Zine Poster](https://github.com/LiamGvchi/gc-minimal-zine-poster).
  95. ## License
  96. MIT. See `LICENSE`.

Ratings and comments

Sign in to rate this skill.

Sign in

Free to use. If it’s helped, pass it on.