agentleFS
full-aigc-pluginsfull-aigc-plugins/dreamina-design-plugin/skills/dreamina-prompt-image2video/SKILL.md

dreamina-prompt-image2video

Skill0 starsChanged 23 days agoApache-2.0View on GitHub

Use when the user has reference image(s) and wants to animate them into a video with 即梦 (Dreamina/Jimeng) Seedance 2.0; mentions "图生视频", "image2video", "图片转视频", "让这张图动起来", "首尾帧", "故事板", "多帧故事", "frames2video"; or supplies images and asks for a video. Covers 4 sub-modes: single image→video (单图生视频), first-last frame transition (首尾帧), multi-frame storyboard (多帧故事), and multi-modal reference compositing (全能参考), with 18 annotated examples, a mode decision tree, and incremental description vocabulary — the reference supplies visuals, the prompt only describes what moves.

1 other project uses a version of this, 1 of them edited.

Adopted by

Most recently taken up by

The file

We split this skill into the rules it gives an agent

6 rules. About 0 are common, 6 are distinctive, and 0 are very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈6

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 0

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Distinctive:

    Use this skill when the user:

    line 13
  2. Distinctive:

    Do NOT use for:

    line 20
  3. Distinctive:

    Ask the user or infer from context:

    line 53
  4. Distinctive:

    Incremental only: Don't repeat what the reference image already shows

    line 95
  5. Distinctive:

    Run through the Validation Checklist.

    line 103
  6. Distinctive:

    Always present: sub-mode label + motion description + complete prompt + duration + camera explanation.

    line 107

≈ “Common” is an estimate: at least 1 in 5 of the 364,980 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

dreamina-prompt-image2video — 即梦图生视频提示词

Craft prompts for animating reference images into videos using Dreamina's 4 video modes.

When to use this skill

Use this skill when the user:

  • Provides image(s) and wants to generate a video from them ("把这张图做成视频")
  • Mentions 图生视频, image2video, 图片转视频, 让图动起来
  • Describes a specific sub-mode: 单图生视频, 首尾帧, 多帧故事, 全能参考
  • Has multiple images and wants to create a narrative video
  • Asks how to describe motion from a static reference image

Do NOT use for:

  • Text-to-video (no reference images) → use dreamina-prompt-text2video
  • CLI execution → use dreamina-cli-image2video
  • Image-to-image editing (output is still an image) → use dreamina-prompt-image2image

Core Methodology

Image-to-video prompting is fundamentally incremental: the reference image already provides the visual content. The prompt only needs to describe what MOVES and CHANGES.

[从哪里开始] + [什么在动/怎么动] + [运镜方式] + [环境/光影变化] + [时长/节奏] + [风格一致]

The I2V Golden Rule

Don't describe what's already in the image. Only describe what happens next.

  • ✗ "一个穿红裙子的女孩站在花园里,阳光灿烂..." — the image already shows this
  • ✓ "女孩缓缓转身面向镜头,红裙随风飘动,镜头慢慢推近" — this is what happens NEXT

4 Sub-Modes Overview

#ModeInputCore Prompt FocusExample File
1单图生视频1 image"从这张图开始,接下来发生什么运动"examples/single-image.md
2首尾帧2 images"如何从图A过渡到图B"examples/first-last-frame.md
3多帧故事2-20 images"帧与帧之间的因果关系和时间流逝"examples/multi-frame.md
4全能参考images+video+audio"综合各参考素材,生成目标视频"examples/multimodal-ref.md

How to use this skill

Step 1: Identify the sub-mode

Ask the user or infer from context:

  • Has 1 image → Sub-mode 1 (单图生视频)
  • Has 2 images described as start/end → Sub-mode 2 (首尾帧)
  • Has 2+ images forming a sequence/storyboard → Sub-mode 3 (多帧故事)
  • Has mixed media (images + video + audio) → Sub-mode 4 (全能参考)

Step 2: Load reference materials

Load the matched example file from examples/ for sub-mode-specific patterns.

Load references/mode-library.md for detailed templates per sub-mode.

Load references/word-library.md for I2V-specific motion and coordination vocabulary.

For camera movement vocabulary, cross-reference the text2video skill's camera-library.md — it covers 7 basic + 6 advanced camera movements with emotional mappings, not duplicated here.

Step 3: Build the prompt (incremental description)

For sub-mode 1 (单图生视频):

  1. Starting point (从哪里开始): Acknowledge the reference image — "从这张画面开始"
  2. What moves (什么在动): The MOST important component. Be specific about WHICH elements move and HOW. "头发在微风中飘动,水面的波纹缓缓扩散" — describe only what changes
  3. What stays still (什么不动): Explicitly state static elements — "建筑和远山保持完全静止"
  4. Camera movement (运镜): How the camera moves through the scene
  5. Duration/pacing (时长/节奏): 4-15 seconds, match complexity
  6. Style consistency (风格一致): The output must match the reference image's style

For sub-mode 2 (首尾帧):

  1. Describe the TRANSITION from frame A to frame B — "花瓣从含苞逐渐绽放至完全盛开"
  2. Describe the pacing of the transition — "前半段缓慢积蓄,后半段加速绽放"
  3. Camera can remain static or follow the transformation

For sub-mode 3 (多帧故事):

  1. Describe the causal relationship between frames — "因为风吹过,所以花瓣飘落"
  2. Describe the passage of time — "从清晨到正午,光影在建筑表面缓缓移动"
  3. Camera work should connect frames narratively

For sub-mode 4 (全能参考):

  1. Assign clear roles to each reference — "图一提供人物特征,图二提供场景风格"
  2. Describe the synthesis — how elements from different references interact

Step 4: Apply I2V-specific rules

  1. Incremental only: Don't repeat what the reference image already shows
  2. Motion specificity: "她的头发在微风中轻轻飘动" not "有风"
  3. Stillness is intentional: Explicitly state what does NOT move
  4. Camera respects the reference: Camera movement is constrained by the reference composition
  5. Style must match: The output must be stylistically consistent with the reference

Step 5: Validate against checklist

Run through the Validation Checklist. Every item must pass before presenting the prompt to the user.

Step 6: Present the result

Always present: sub-mode label + motion description + complete prompt + duration + camera explanation.

Writing Rules for Image-to-Video

Rule 1: Incremental description — only write what changes

The reference image IS the visual. Don't describe its static content in the prompt. Focus 100% on motion and change.

Rule 2: Explicit stillness is as important as motion

Without explicit stillness constraints, AI tends to animate everything. Write "保持建筑完全静止""只有水在流动,其他一切不动".

Rule 3: Match sub-mode to prompt structure

Each sub-mode needs a different prompt structure. Don't use a single-image prompt for a frames2video task.

Rule 4: Camera movement respects the reference

Unlike text-to-video where camera can go anywhere, I2V camera movement is bound by the reference image's composition. Camera descriptions should work WITH the existing composition.

Rule 5: Duration scales with complexity

Sub-modeMotion ComplexitySuggested Duration
单图1-2 moving elements5-8s
首尾帧1 complete transformation8-12s
多帧Multi-stage narrative10-15s
全能参考Complex multi-source synthesis8-12s

Validation Checklist

  • Prompt describes only what MOVES/CHANGES (not static content from the reference image)
  • Specific elements are named as moving + specific motion verbs used
  • Elements that should stay still are explicitly listed
  • Camera movement is described (not just named — explain the movement logic)
  • Duration matches motion complexity
  • Style consistency with reference image is addressed
  • Correct sub-mode is identified and communicated

Gotchas

  1. Don't describe the reference image — the #1 I2V mistake. "A girl in a red dress standing in a garden" is useless if the reference already shows this. Write what HAPPENS: "The girl slowly turns..."
  2. Everything moves unless you stop it — AI's default is to animate everything. Use "everything else remains completely still" to lock down static elements
  3. Camera terms need explanation, not just naming — 只写"推镜头"不够,需要写"镜头缓缓向前推进,逐渐聚焦人物的面部表情"
  4. Resolution varies by sub-mode — Seedance 2.0 = 720p, legacy 3.5pro frames2video = 1080p
  5. Multiframe limit — 2-20 images for multi-frame story mode. More images = better consistency but slower generation
  6. Multimodal limits — max 9 images + 3 videos + 3 audio clips
  7. Style drift possible — output video may not perfectly match reference image style; use "保持参考图的风格和色调完全一致" to minimize drift

Available Resources

ResourceDescriptionWhen to Load
references/mode-library.md4 sub-mode templates with prompt structuresWhen building any I2V prompt
references/word-library.mdI2V-specific motion, stillness, coordination vocabularyWhen building any I2V prompt
examples/single-image.md单图生视频 examplesUser has 1 reference image
examples/first-last-frame.md首尾帧 examplesUser has 2 images as start/end
examples/multi-frame.md多帧故事 examplesUser has 2+ sequential images
examples/multimodal-ref.md全能参考 examplesUser has mixed media references
<!-- QUALITY_BASELINE_V1 -->

When to use(什么时候使用)

当用户需要 在明确输入、预算和交付约束后执行生成或写入操作 时加载本技能。先从请求中提取目标、输入、约束、交付格式和验收标准;描述摘要为:Use when the user has reference image(s) and wants to animate them into a video with 即梦 (Dreamina/Jimeng) Seedance 2.0; mentions "图生视频", "image2video", "图片转视频", "让这张图动起来", "首尾帧", "故事板", "多帧故事", "frames2video"; or supplies images and asks for a video. Covers 4 sub-modes: single image→video (单图生视频), first-last frame transition (首尾帧), multi-frame storyboard (多帧故事), and multi-modal reference compositing (全能参考), with 18 annotated examples, a mode decision tree, and incremental description vocabulary — the reference supplies visuals, the prompt only describes what moves.。

Rules

  • 先读后写:先确认当前状态与真实能力,再执行会改变外部状态的动作。
  • 权限最小化:只使用完成当前步骤所需的文件、工具、账户与网络范围。
  • 证据优先:运行结果、资源 ID、版本、哈希或测试输出缺失时,明确标记为 NOT_VERIFIED。
  • 幂等优先:保留请求标识与阶段状态;结果不明确时先查询,不进行盲目重试。
  • 隐私安全:日志、示例、回执和错误信息不得包含 token、cookie、密钥或个人敏感数据。

Workflow

Step 1:澄清意图

确认本技能是否匹配目标;若只是相邻需求,交给更精确的技能。

Step 2:执行预检

校验输入、模型/工具能力、输出路径、预算上限和审批状态;任一关键条件未知时停止在只读阶段。

Step 3:形成计划

列出将调用的工具、会改变的对象、成功标准以及失败后的安全退出方式。

Step 4:执行动作

按一次批准执行并记录请求标识;模糊结果先查询而不是重提;每个外部调用均保留可关联的状态或回执。

Step 5:验证交付

验证产物存在性、格式、哈希/标识、成本状态和质量门禁,并把事实、推断和未验证项分开陈述。

Validation checklist

  • 技能触发条件与用户意图一致,没有把相邻任务误路由到本技能。
  • 输入、目标对象、版本和输出位置均已明确,且没有使用猜测值替代必填值。
  • 所有写入、付费、发布或不可逆动作都在用户授权范围内。
  • 结果已用独立检查验证;仅有“命令成功”或“文件存在”不算完整验收。
  • 输出包含实际证据、失败/跳过项、剩余风险和可执行的下一步。

Gotchas

  1. 把计划当结果:文档或提示词不等于真实执行;必须标明实际运行层级。
  2. 错误重试:超时或响应丢失可能已经产生远端状态,先查询再决定是否重试。
  3. 隐式扩大范围:批量、全量、发布、覆盖和付费不是普通读写的自然延伸。
  4. 版本漂移:引用外部资源时记录版本、tag 或提交;不要把可变分支当发布证据。
  5. 证据过期:缓存、旧截图和历史测试不能证明当前环境;在交付前刷新关键证据。

不适用与边界

付费、发布、覆盖、上传或外部写入必须使用当前任务的显式授权;不自动扩大次数和预算。 如果请求需要别的技能,不复制其正文;按技能名称进行交接,并保留当前任务上下文。

Progressive disclosure

  • 需要确定输入/输出、状态和授权点时,读取 references/workflow-contract.md。
  • 需要交付前自检时,读取 references/validation-checklist.md。
  • 遇到超时、部分成功或恢复场景时,读取 references/error-recovery.md。
  • 首次运行、拒绝越权和失败恢复分别参考 examples/happy-path.md、examples/boundary-refusal.md、examples/failure-recovery.md。

Download now (ZIP)

Does it work?

Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.

No reviews yet

People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.

  • It worked
  • Partly
  • It didn't
What did you use it for? e.g. Reviewing pull requests in a Rails app

Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.

Your trail

Files you open will collect here, so you can find them again.

Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.

Now do this for your own.

Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.