video-generation

Compare original and translation side by side

🇺🇸

Original

English

🇨🇳

Translation

Chinese

Video Generation

视频生成

Use When

使用场景

The desired final asset is a generated video or video render plan.
The request may be a simple clip, storyboard, UGC, podcast clip, reference video, talking-head, image-to-video, text-to-video, or research handoff.
The next decision is task class, request architecture, and execution handoff.

目标最终资产为生成的视频或视频渲染方案。
请求类型可能为简单片段、分镜脚本、UGC、播客片段、参考视频、访谈类视频、图生视频、文生视频或研究交付内容。
下一步需要确定任务类别、请求架构并完成执行交付。

Do Not Use When

禁用场景

The user only needs transcription, subtitles, semantic analysis, or edit decisioning. Use
```
media-router
```
.
The request architecture is already approved and provider-ready. Use
```
seedance-submitter
```
or
```
video-batch-runner
```
.
The task is image-only or audio-only.

用户仅需要转录、字幕、语义分析或编辑决策时，请使用
```
media-router
```
。
请求架构已获批准且可直接提交给服务商时，请使用
```
seedance-submitter
```
或
```
video-batch-runner
```
。
任务仅涉及图片或音频内容时。

Core Boundary

核心边界

This is the video generation controller. It does not submit jobs.

It must:

classify the video task,
choose whether references or storyboard must be resolved,
hand approved structure to
```
video-request-architect
```
,
hand submit-ready plans to
```
seedance-submitter
```
or
```
video-batch-runner
```
.

本工具为视频生成控制器，不负责提交任务。

它必须：

对视频任务进行分类，
判断是否需要解析参考内容或分镜脚本，
将已批准的结构交付给
```
video-request-architect
```
，
将可直接提交的方案交付给
```
seedance-submitter
```
或
```
video-batch-runner
```
。

Task Classes

任务类别

Task class	Use when	Required handoff
`simple_clip`	one short text/image-to-video clip	`video-request-architect`
`storyboard`	panel sequence or contact sheet drives the clip	`storyboard-grid-writer` , then `video-request-architect`
`ugc`	creator-style product or app video	`ugc-flow` unless it is one isolated clip
`podcast`	voice-led clip, audiogram, or speaker segment	`audio-generation` , then video architecture
`reference_video`	benchmark video controls rhythm or motion	`reference-decode` and `reference-contract-builder`
`research_handoff`	research/benchmark becomes a generation request	`benchmark-to-brief` , then `video-request-architect`

任务类别	使用场景	需交付至
`simple_clip`	单个短时长文生/图生视频片段	`video-request-architect`
`storyboard`	由分镜序列或样片表驱动的视频片段	`storyboard-grid-writer` ，随后交付至 `video-request-architect`
`ugc`	创作者风格的产品或应用视频	若为独立片段则直接使用 `ugc-flow`
`podcast`	语音主导的片段、音频波形图或演讲者片段	`audio-generation` ，随后进行视频架构设计
`reference_video`	由基准视频控制节奏或运动的视频	`reference-decode` 和 `reference-contract-builder`
`research_handoff`	将研究/基准内容转化为生成请求	`benchmark-to-brief` ，随后交付至 `video-request-architect`

Routing Table

路由表

If not video-generation	Send to
Understanding or subtitles first	`media-router`
Image asset must be created first	`image-generation`
Voice or dub must be created first	`audio-generation`
Product-to-UGC pipeline	`ugc-flow`
Provider-ready request exists	`seedance-submitter` or `video-batch-runner`

非视频生成场景	发送至
需先进行内容理解或字幕处理	`media-router`
需先创建图片资产	`image-generation`
需先创建语音或配音内容	`audio-generation`
产品转UGC流水线	`ugc-flow`
存在可直接提交给服务商的请求	`seedance-submitter` 或 `video-batch-runner`

Three-Layer Boundary

三层边界

Controller: classify task and produce the handoff. It must not submit.
Architect:
```
video-request-architect
```
builds provider-agnostic structure. It must not call providers.
Runner:
```
seedance-submitter
```
or
```
video-batch-runner
```
validates and executes normalized requests. It must not make creative strategy decisions.

控制器：分类任务并完成交付，不得提交任务。
架构师：
```
video-request-architect
```
构建与服务商无关的结构，不得调用服务商接口。
执行器：
```
seedance-submitter
```
或
```
video-batch-runner
```
验证并执行标准化请求，不得制定创意策略决策。

Output Shape

输出格式

Return:

```
taskClass
```
```
requiredUpstreamSkills
```
```
architectureHandoff
```
```
runnerHandoff
```
```
referencePolicy
```
```
mustNotDo
```

返回内容需包含：

```
taskClass
```
```
requiredUpstreamSkills
```
```
architectureHandoff
```
```
runnerHandoff
```
```
referencePolicy
```
```
mustNotDo
```

Fail Fast

快速失败机制

Stop if duration, aspect ratio, source basis, product identity, required references, audio needs, or storyboard structure would change the route.
Do not invent a provider payload before the architecture is coherent.
Do not ask a runner to infer hook, persona, or reference meaning.

若时长、宽高比、来源基础、产品标识、所需参考内容、音频需求或分镜结构会改变路由方向，立即停止流程。
在架构未明确前，不得生成服务商请求 payload。
不得要求执行器推断钩子、角色或参考内容的含义。

Public Command Boundary

公开命令边界

Check readiness first:

postplus doctor --skill video-generation

This public skill is instruction-driven. Produce the controller handoff artifact directly from the available evidence.
Do not call private provider/runtime paths or unpublished local tools.
If the CLI returns a quote-confirmation challenge, run
```
postplus quote confirm --json --challenge-file <challenge.json>
```
and retry with the returned token.

先检查就绪状态：

postplus doctor --skill video-generation

。

本公开技能由指令驱动，需直接根据现有证据生成控制器交付产物。
不得调用私有服务商/运行时路径或未发布的本地工具。
若CLI返回报价确认挑战，请运行
```
postplus quote confirm --json --challenge-file <challenge.json>
```
并使用返回的令牌重试。