ra-video-production-director

Compare original and translation side by side

🇺🇸

Original

English

🇨🇳

Translation

Chinese

AI 视频创作 Skill

Role

角色

Act as the director layer above the existing video skills. Route the job, define the acceptance standard, then coordinate production and QC. Do not replace

hyperframes

remotion

ra-audio-to-subtitles

skill-captions

embedded-captions

graphic-overlays

ra-video-download

视频转写

tts-skill

heygen-digital-avatar

, or

ai-video-qc-delivery

; load those skills only when their lane is selected.

If the user provides a source video URL or local file and asks for

视频洗稿

视频二创

, or

洗稿并制作视频

, route to

ra-video-wash-pipeline

before this skill so extraction and rewrite happen first.

在现有视频 Skill 之上充当导演层。负责路由任务、定义验收标准，然后协调制作与 QC。不要替代

hyperframes

、

remotion

、

ra-audio-to-subtitles

、

skill-captions

、

embedded-captions

、

graphic-overlays

、

ra-video-download

、

视频转写

、

tts-skill

、

heygen-digital-avatar

或

ai-video-qc-delivery

；仅在选中相应工作流时加载这些 Skill。

如果用户提供源视频 URL 或本地文件，并要求

视频洗稿

、

视频二创

或

洗稿并制作视频

，应先路由到

ra-video-wash-pipeline

，完成提取和改写后，再使用本 Skill。

待制作 Queue Intake

待制作队列接入

If the input is a handoff project folder under

01-内容生产/视频工作台/待制作/<日期-主题>/

(contract file

交接稿.md

at its root; or the user says

制作待制作队列

按交接稿制作

, or points at a queued script):

the 交接稿.md frontmatter is the binding production contract — platform, ratio,
```
duration_target_s
```
,
```
word_count
```
,
```
voice
```
,
```
caption_style
```
,
```
visual
```
,
```
illustration_skill
```
,
```
typography
```
override intake defaults and must never be replaced by an execution skill's own defaults (e.g. HyperFrames TTS voices are not a substitute for the contract voice)
before starting work, set
```
status: 制作中
```
in 交接稿.md and move the whole project folder to
```
01-内容生产/视频工作台/制作中/<日期-主题>/
```
, so the other machine does not double-produce; all engineering happens inside that folder
the script segments in the handoff file are the scene units; do not re-split or re-flow the narration
skip intake questions the contract already answers

如果输入是位于

01-内容生产/视频工作台/待制作/<日期-主题>/

下的交接项目文件夹（根目录有合同文件

交接稿.md

；或者用户说

制作待制作队列

、

按交接稿制作

，或指向排队中的脚本）：

```
交接稿.md
```
的 frontmatter 是具有约束力的制作合同——平台、比例、
```
duration_target_s
```
、
```
word_count
```
、声音、
```
caption_style
```
、
```
visual
```
、
```
illustration_skill
```
、
```
typography
```
会覆盖接入默认值，且绝不能被执行 Skill 自身的默认值替换（例如，HyperFrames TTS 声音不能替代合同指定的声音）
开始工作前，在
```
交接稿.md
```
中设置
```
status: 制作中
```
，并将整个项目文件夹移动到
```
01-内容生产/视频工作台/制作中/<日期-主题>/
```
，避免其他机器重复制作；所有工程工作都在该文件夹内进行
交接文件中的脚本分段是场景单元；不得重新拆分或重新编排旁白
合同已经回答的问题无需再次询问

Hard Rules

硬性规则

Define the output standard before production: original video, pixel-level alignment, visual replica, style-inspired creation, captioned footage, narrated explainer, product promo, website tour, or motion graphic.
Do not claim pixel-level alignment for a hand-authored HyperFrames/Remotion rebuild. Pixel-level requires source-stream reuse, original project assets, or hard metrics such as matching hash,
```
cmp
```
, PSNR infinity, or SSIM 1.0.
Keep source projects and intermediates under
```
<WORKSPACE>/01-内容生产/视频工作台/制作中/
```
; the workbench
```
01-内容生产/视频工作台/
```
holds only handoffs, transcripts, and archives.
Final deliverables stay inside the vault. Every video produced by this system — queue or non-queue, transcript-direct, practical long-form, translated/repackaged, or ad hoc — delivers to
```
01-内容生产/视频工作台/已制作/<月上旬|月下旬>/<日期-主题>/成片/
```
, with key QC images in the sibling
```
质检/
```
. Use the delivery date: days 1-15 go to
```
X月上旬
```
and days 16 through month-end go to
```
X月下旬
```
.
```
已制作/6月历史成片库/
```
（封存只读） and the desktop
```
<WORKSPACE>/legacy-archive/
```
are retired legacy history — do not write new deliverables there.
Publish titles are not production work for 待制作 queue jobs:
```
ra-video-title
```
already wrote 8-12 candidates into the handoff 标题候选 section at wash time, and the final title is chosen manually by the user at publish time. Do not generate or apply publish titles and do not create
```
视频标题.md
```
for queue jobs (on-page scene titles from the script are unaffected). Only for non-queue jobs (no handoff contract) generate 8-12 candidates with
```
ra-video-title
```
after QC and let the user choose before saving
```
视频标题.md
```
next to the final MP4.
A video is not complete until media specs, visual checks, and archive paths are confirmed. The selected-title-file requirement applies to non-queue jobs only; queue jobs finish without
```
视频标题.md
```
(title selection stays manual with the user).
If the user explicitly names a visual or illustration skill, load and follow that skill before designing the visual layer. Do not replace a requested illustration skill with hand-coded shapes, CSS drawings, or "style-like" approximations unless the user explicitly approves that substitution.
When full AI narration is required, final deliverables must use local IndexTTS2 through
```
tts-skill
```
, voice id
```
pluvio-indextts2-calm-v1
```
, and the canonical lossless WAV declared in
```
automation/config/tts-routing.json
```
. MiniMax is reserved for public article/tutorial relay examples and is forbidden for local video narration. Do not silently fall back to macOS
```
say
```
, browser speech, MiniMax, or another generic/cloud voice. If the local model or canonical WAV is missing, repair the IndexTTS2 lane or stop before final render.
For jobs taken from the
```
待制作
```
queue, delivery is not complete until
```
python3 automation/scripts/check_delivery.py <交接文件> <项目目录> <成片.mp4>
```
exits 0, the 制作端回执 in the handoff file is filled (paths + checker output),
```
status
```
is set to
```
已完成
```
, and the per-project archive folder
```
01-内容生产/视频工作台/已制作/<月上旬|月下旬>/<日期-主题>/
```
(folder name = the project folder name) is assembled by moving the cleaned project folder from
```
制作中/
```
:
```
交接稿.md
```
at its root, the final MP4 in
```
成片/
```
, key QC frames in
```
质检/
```
, minimal engineering in
```
工程/
```
. Non-queue jobs assemble the same dated archive structure using a clear topic slug; copy a production note/script/plan into the archive root when available. No job copies a new final into
```
已制作/6月历史成片库/
```
（封存只读）.
The 图文工作台 queue follows the same contract pattern: take
```
01-内容生产/图文工作台/待制作/<file>.md
```
(frontmatter contract: card_count, ratio, style, illustration_skill, image_skill), set
```
status: 制作中
```
on start, produce the image cards without rewriting the copy and without generating or applying publish titles, deliver to
```
01-内容生产/图文工作台/已制作/<日期-主题>/
```
(handoff file moved in + finished cards under
```
图/
```
), set
```
status: 已完成
```
, and register the item under 待修改 in
```
01-内容生产/发布看板.md
```
. The source-privacy wall applies unchanged.
实操长片 projects follow the unified engineering layout under
```
01-内容生产/视频工作台/制作中/<日期-主题>/
```
; do not create a separate planning directory under
```
视频工作台/
```
. the 策划稿 frontmatter
```
status
```
is the only handoff signal, and the user handles 录制 and 剪映融合 stages personally. Codex owns three stages: (a)
```
status: 待粗剪
```
— default to
```
ra-local-talking-head-cut
```
: Seed ASR word timestamps, terminology/uncertainty review, and a source-timeline pre-cut SRT. Stop for explicit user approval; no EDL or render is allowed before approval. Then perform semantic cleanup, adaptive pauses, source-size/source-fps rendering, dialogue loudness, media QC, and final- media subtitle generation with caption QC PASS. Use
```
video-use
```
as the multiple-take evidence/selection layer; use
```
Agentchengfeng/chengfeng-videocut-skills
```
only for word-level web review or an intentionally aggressive comparison; choose
```
AI剪口播
```
when the user wants waveform review and an FCPXML timeline for 剪映/Final Cut Pro. The default MP4 workflow delivers
```
粗剪/口播-粗剪.mp4
```
and sets
```
status: 融合中
```
; the FCPXML workflow moves forward only after the editor timeline is accepted; (b)
```
status: 待字幕
```
— run
```
ra-audio-to-subtitles
```
against the user's final merged cut, require word-timestamp QC to pass, then run
```
skill-captions
```
with the contract
```
caption_style
```
(default
```
anchor-dark
```
) to preview, burn, and validate the phrase captions; save the matching
```
.srt
```
and passing
```
caption-render-qc.json
```
, then set
```
status: 待动画
```
; (c)
```
status: 待动画
```
— add Remotion component overlays ONLY to the 全身出镜 segments listed in the 策划稿 frontmatter, composite, deliver the final MP4 to
```
视频工作台/已制作/<月上旬|月下旬>/<日期-主题>/成片/
```
, put key QC in the sibling
```
质检/
```
, register on 发布看板, set
```
status: 成片
```
. All rough-cut tools are rough-cut-only in this workspace: never use any rough- cut tool's subtitle output as the production subtitle timeline.
For any narrated final video, the final concatenated audio or final merged video is the only subtitle timing source. Run
```
ra-audio-to-subtitles
```
after audio is locked; archive
```
captions_words.json
```
,
```
captions.json
```
,
```
captions.srt
```
, and
```
caption-qc.json
```
. Character-count interpolation, scene-duration interpolation, and fixed-delay caption timing are preview- only and must never enter a recommended final.
Whenever a video component, video background, or reference-video replica reaches a rendered, QC-passed final, stop before adding it to
```
05-视频组件/
```
and explicitly ask the user: whether it should enter the component library, which existing category should contain it, and whether a new category is needed. Do not infer placement or create a category before the user answers. If the user declines, use only the normal finished-video archive. If the user approves, treat the confirmed
```
05-视频组件/<中文分类>/<中文组件名>/
```
names as canonical and make the component directory self-contained in the same task: portable MP4,
```
manifest.json
```
,
```
质检/
```
, and
```
工程/
```
(the sole editable source, stripped of node_modules/renders). No engineering mirrors, symlinks, or second copies exist anywhere else; register the component in
```
05-视频组件/CATALOG.md
```
before reporting completion.
Whenever the user asks for a HeyGen/digital-human/avatar video, or a video production contract calls for the existing Pluviobyte digital human, load
```
heygen-digital-avatar
```
before generation. Its original Digital Twin look, CLI OAuth subscription billing, Avatar III default, local IndexTTS2 audio, one-master reuse, and approved
```
headroom_08
```
lower-left circle layout are binding. Only an explicit request for
```
全屏数字人
```
may bypass the circle layout.
Treat any decorative focus frame enclosing over 40% of the canvas or multiple content groups as an entrance cue only. Its total visibility must not exceed the shorter of 2 seconds or 15% of the scene, and it must exit before the steady information state. Persistent frames require explicit semantic purpose and must tightly bound one local target.

制作前必须定义输出标准：原创视频、像素级对齐、视觉复刻、风格启发式创作、带字幕的实拍素材、解说型说明视频、产品宣传片、网站导览或动态图形。
对手工编写的 HyperFrames/Remotion 重建版本，不得声称达到像素级对齐。像素级对齐要求复用源视频流、原始项目素材，或具备匹配哈希、
```
cmp
```
、PSNR infinity 或 SSIM 1.0 等硬性指标。
源项目和中间产物必须保存在
```
<WORKSPACE>/01-内容生产/视频工作台/制作中/
```
下；工作台
```
01-内容生产/视频工作台/
```
仅用于存放交接文件、转写稿和归档内容。
最终交付物必须保存在 vault 内。该系统制作的每个视频——无论是队列任务、非队列任务、转写稿直制、实操长片、翻译/重新打包还是临时任务——都应交付到
```
01-内容生产/视频工作台/已制作/<月上旬|月下旬>/<日期-主题>/成片/
```
，关键 QC 图像放在同级
```
质检/
```
目录中。根据交付日期归档：1 至 15 日进入
```
X月上旬
```
，16 日至月末进入
```
X月下旬
```
。
```
已制作/6月历史成片库/
```
（封存只读）和桌面目录
```
<WORKSPACE>/legacy-archive/
```
是已退役的历史归档，不得在其中写入新的交付物。
对
```
待制作
```
队列任务，发布标题不属于制作工作：
```
ra-video-title
```
已在洗稿时将 8 至 12 个候选标题写入交接文件的
```
标题候选
```
部分，最终标题由用户在发布时手动选择。不得生成或应用发布标题，也不得为队列任务创建
```
视频标题.md
```
（脚本中的页面场景标题不受影响）。仅对于没有交接合同的非队列任务，在 QC 后使用
```
ra-video-title
```
生成 8 至 12 个候选标题，让用户选择后，再将
```
视频标题.md
```
保存到最终 MP4 旁边。
在确认媒体规格、视觉检查和归档路径之前，视频不得视为完成。标题文件要求仅适用于非队列任务；队列任务无需
```
视频标题.md
```
，标题仍由用户手动选择。
如果用户明确指定视觉或插画 Skill，必须先加载并遵循该 Skill，再设计视觉层。除非用户明确同意替换，否则不得使用手写形状、CSS 绘图或“style-like”近似方案替代指定的插画 Skill。
需要完整 AI 旁白时，最终交付物必须通过
```
tts-skill
```
使用本地 IndexTTS2，声音 ID 为
```
pluvio-indextts2-calm-v1
```
，并使用
```
automation/config/tts-routing.json
```
中声明的标准无损 WAV。MiniMax 仅用于公开文章或教程转述示例，禁止用于本地视频旁白。不得在未说明的情况下回退到 macOS
```
say
```
、浏览器语音、MiniMax 或其他通用/云端声音。如果本地模型或标准 WAV 缺失，应修复 IndexTTS2 工作流，或在最终渲染前停止。
对于来自
```
待制作
```
队列的任务，必须确保
```
python3 automation/scripts/check_delivery.py <交接文件> <项目目录> <成片.mp4>
```
以状态码 0 退出、在交接文件中填写
```
制作端回执
```
（路径和检查器输出）、将
```
status
```
设置为
```
已完成
```
，并按以下方式组装项目归档文件夹
```
01-内容生产/视频工作台/已制作/<月上旬|下旬>/<日期-主题>/
```
：文件夹名称必须与项目文件夹名称一致，通过移动
```
制作中/
```
中清理后的项目文件夹完成归档；根目录保留
```
交接稿.md
```
，最终 MP4 位于
```
成片/
```
，关键 QC 帧位于
```
质检/
```
，必要的精简工程文件位于
```
工程/
```
。非队列任务使用清晰的主题 slug 构建相同的日期归档结构；如果存在制作说明、脚本或计划，将其复制到归档根目录。任何任务都不得将新的最终成片复制到
```
已制作/6月历史成片库/
```
（封存只读）。
图文工作台队列遵循相同的合同模式：接收
```
01-内容生产/图文工作台/待制作/<file>.md
```
（frontmatter 合同字段包括
```
card_count
```
、
```
ratio
```
、
```
style
```
、
```
illustration_skill
```
、
```
image_skill
```
），开始时将
```
status
```
设置为
```
制作中
```
，生成图文卡片时不得改写文案，也不得生成或应用发布标题；交付到
```
01-内容生产/图文工作台/已制作/<日期-主题>/
```
（移动交接文件，并将完成卡片放入
```
图/
```
），将
```
status
```
设置为
```
已完成
```
，并在
```
01-内容生产/发布看板.md
```
的
```
待修改
```
下登记该项目。源素材隐私隔离墙规则保持不变。
实操长片项目遵循
```
01-内容生产/视频工作台/制作中/<日期-主题>/
```
下的统一工程目录结构；不得在
```
视频工作台/
```
下另建规划目录。
```
策划稿
```
frontmatter 中的
```
status
```
是唯一的交接信号，录制和剪映融合阶段由用户亲自处理。Codex 负责以下三个阶段： (a)
```
status: 待粗剪
```
——默认使用
```
ra-local-talking-head-cut
```
：生成 Seed ASR 词级时间戳，进行术语/不确定性检查，并基于源时间线生成粗剪 SRT。必须停止并等待用户明确批准；批准前不得生成 EDL 或渲染。随后执行语义清理、自适应停顿、按源尺寸/源帧率渲染、对白响度处理、媒体 QC，并基于最终媒体生成字幕且字幕 QC 必须为 PASS。使用
```
video-use
```
作为多条素材的证据/选择层；仅在需要逐词网页审阅或有意的强化对比时使用
```
Agentchengfeng/chengfeng-videocut-skills
```
；当用户需要波形审阅以及用于剪映/Final Cut Pro 的 FCPXML 时间线时，选择
```
AI剪口播
```
。默认 MP4 工作流交付
```
粗剪/口播-粗剪.mp4
```
，并将
```
status
```
设置为
```
融合中
```
；FCPXML 工作流仅在剪辑时间线通过验收后继续； (b)
```
status: 待字幕
```
——针对用户最终合并的剪辑运行
```
ra-audio-to-subtitles
```
，要求词级时间戳 QC 通过，然后使用合同中的
```
caption_style
```
（默认
```
anchor-dark
```
）运行
```
skill-captions
```
，用于预览、烧录和验证短语字幕；保存匹配的
```
.srt
```
和通过检查的
```
caption-render-qc.json
```
，然后将
```
status
```
设置为
```
待动画
```
； (c)
```
status: 待动画
```
——仅向
```
策划稿
```
frontmatter 中列出的
```
全身出镜
```
片段添加 Remotion 组件叠加层，完成合成后，将最终 MP4 交付到
```
视频工作台/已制作/<月上旬|月下旬>/<日期-主题>/成片/
```
，把关键 QC 内容放入同级
```
质检/
```
，在
```
发布看板
```
中登记，并将
```
status
```
设置为
```
成片
```
。本工作区中的所有粗剪工具都只能用于粗剪：绝不能使用任何粗剪工具生成的字幕输出作为正式制作字幕时间线。
对任何带旁白的最终视频，最终拼接音频或最终合并视频是唯一的字幕时间来源。音频锁定后必须运行
```
ra-audio-to-subtitles
```
，并归档
```
captions_words.json
```
、
```
captions.json
```
、
```
captions.srt
```
和
```
caption-qc.json
```
。按字符数插值、按场景时长插值以及固定延迟字幕计时仅可用于预览，绝不能进入推荐的最终成片。
每当视频组件、视频背景或参考视频复刻版本完成渲染并通过 QC 时，在将其加入
```
05-视频组件/
```
前必须停止，并明确询问用户：是否应进入组件库、应归入哪个现有分类，以及是否需要新建分类。在用户回答前，不得推断存放位置或创建分类。如果用户拒绝，则仅使用正常的成片归档。如果用户批准，则将确认后的
```
05-视频组件/<中文分类>/<中文组件名>/
```
名称视为标准名称，并在同一任务中使组件目录自包含：包含可移植 MP4、
```
manifest.json
```
、
```
质检/
```
和
```
工程/
```
（唯一的可编辑源，并移除 node_modules/renders）。其他任何位置都不得存在工程镜像、符号链接或第二份副本；报告完成前，必须在
```
05-视频组件/CATALOG.md
```
中登记该组件。
每当用户要求制作 HeyGen、数字人或 Avatar 视频，或者视频制作合同要求使用现有的 Pluviobyte 数字人时，必须在生成前加载
```
heygen-digital-avatar
```
。其原始 Digital Twin 外观、CLI OAuth 订阅计费、Avatar III 默认设置、本地 IndexTTS2 音频、one-master 复用机制以及已批准的
```
headroom_08
```
左下角圆形布局均具有约束力。只有用户明确要求
```
全屏数字人
```
时，才可以绕过圆形布局。
任何包围画布超过 40% 或包含多个内容组的装饰性焦点框，都只能作为入场提示。其总可见时长不得超过 2 秒或场景时长的 15%，两者取较短者，并且必须在画面进入稳定的信息状态前退出。持续存在的边框必须具有明确的语义用途，并紧密包围一个局部目标。

Default Visual System

默认视觉系统

When the user says to make a video with this skill and does not provide a conflicting visual direction, use a white-first visual system:

background: pure white or warm white, with generous blank space
for white-first AI/tool tutorials, page-based explainers, and white motion graphics, read
```
references/perspective-grid-v5.md
```
and start from the workspace component
```
05-视频组件/视频背景/透视网格背景/
```
. The bundled
```
assets/perspective-grid-v5/
```
copy is the same-hash Skill mirror and fallback, not a second independently editable baseline. Use the approved V5 loop as the bottom-most visual layer at its native playback rate: 1920x1080, 30fps, 7.3 seconds / 219 frames, three complete grid cells per loop, and a 2.433-second single-cell period. Preserve its geometry, asymmetric color plate, center fog, phase, and seamless-loop timing. Do not rebuild it from an older version or an ad-hoc CSS grid.
disable the V5 grid only when source footage, full-screen talking head, dense UI, readability, a handoff contract, or an explicit reference conflicts with it. Record the reason in the production note.
primary marks: black text, black line art, and restrained dark UI elements
accents: use only small amounts of red, orange, or blue for emphasis, warnings, arrows, feedback, or state
captions: for narrated explainers over footage, screenshots, the V5 grid, or variable backgrounds, run
```
skill-captions
```
and use its
```
anchor-dark
```
shrink-wrapped panel by default. A handoff
```
caption_style
```
or explicit user style overrides it; frameless captions require a controlled uniform light caption zone and a recorded reason.
avoid defaulting to dark, purple, neon, or SaaS-magic styling unless the user explicitly asks for it or the selected route requires it
use
```
ian-xiaohei-illustrations
```
as the default illustration source for page-based or scene-based videos. Prefer the active project-local copies under
```
.agents/skills/ian-xiaohei-illustrations/
```
,
```
.codex/skills/ian-xiaohei-illustrations/
```
, and
```
.claude/skills/ian-xiaohei-illustrations/
```
; the shared library path
```
<WORKSPACE>/.claude/skills/ian-xiaohei-illustrations
```
is only a source of truth to repair from when the project copy is missing.
give every page or scene a meaningful Xiaohei illustration whenever possible; skip only when footage, screenshots, dense UI, or readability would clearly be harmed
each illustration must explain the page's core claim, process, contrast, or state; Xiaohei should perform the core action, not appear as decoration
if the plan says "Xiaohei" or "小黑配图", create real image assets with
```
ian-xiaohei-illustrations
```
and save them into the project. A CSS/HTML approximation of Xiaohei does not satisfy this requirement.
for page-based 16:9 Xiaohei videos, use the
```
xiaohei-16x9-layout
```
contract before writing the composition. The contract fixes the page skeleton first: left-top title, left-middle note card, right-middle Xiaohei illustration, and lower caption. Do not invent a fresh layout unless the user asks for a different visual system.

当用户要求使用本 Skill 制作视频且未提供相冲突的视觉方向时，采用白色优先的视觉系统：

背景：纯白或暖白色，并保留充足的留白
对于白色优先的 AI/工具教程、基于页面的说明视频和白色动态图形，读取
```
references/perspective-grid-v5.md
```
，并从工作区组件
```
05-视频组件/视频背景/透视网格背景/
```
开始构建。内置的
```
assets/perspective-grid-v5/
```
副本是同哈希的 Skill 镜像和备用版本，不是第二套独立可编辑基线。将已批准的 V5 循环以原生播放速率作为最底层视觉层：1920x1080、30fps、7.3 秒/219 帧、每个循环包含三个完整网格单元，单个单元周期为 2.433 秒。必须保留其几何结构、非对称色块、中心雾化效果、相位和无缝循环时序。不得基于旧版本或临时 CSS 网格重新构建。
仅当源视频、全屏出镜、密集 UI、可读性、交接合同或明确参考与此冲突时，才停用 V5 网格。必须在制作说明中记录原因。
主要视觉元素：黑色文字、黑色线稿，以及克制的深色 UI 元素
强调色：仅少量使用红色、橙色或蓝色，用于强调、警告、箭头、反馈或状态展示
字幕：对于叠加在实拍素材、截图、V5 网格或可变背景上的解说型说明视频，运行
```
skill-captions
```
，默认使用其
```
anchor-dark
```
自动收缩包裹面板。交接文件中的
```
caption_style
```
或用户明确指定的样式具有更高优先级；无边框字幕需要使用受控、均匀的浅色字幕区域，并记录原因。
除非用户明确要求，或所选工作流必须如此，否则不要默认使用深色、紫色、霓虹或 SaaS 魔法风格

将

ian-xiaohei-illustrations

作为基于页面或场景的视频的默认插画来源。优先使用当前项目下的副本：

.agents/skills/ian-xiaohei-illustrations/

、

.codex/skills/ian-xiaohei-illustrations/

和

.claude/skills/ian-xiaohei-illustrations/

；共享库路径

<WORKSPACE>/.claude/skills/ian-xiaohei-illustrations

仅作为项目副本缺失时用于修复的来源

尽可能为每个页面或场景提供有意义的 Xiaohei 插画；只有实拍素材、截图、密集 UI 或可读性明显会受到影响时才可跳过
每幅插画都必须解释页面的核心论点、流程、对比关系或状态；Xiaohei 应执行核心动作，而不是仅作为装饰出现
如果方案中写有“Xiaohei”或“小黑配图”，必须使用
```
ian-xiaohei-illustrations
```
创建真实图像资产并保存到项目中。CSS/HTML 绘制的 Xiaohei 近似图不满足此要求
对于基于页面的 16:9 Xiaohei 视频，在编写构图之前使用
```
xiaohei-16x9-layout
```
合同。该合同首先固定页面骨架：左上标题、左中说明卡片、右中 Xiaohei 插画和底部字幕。除非用户要求采用其他视觉系统，否则不得自行创造全新布局

Page-Based Explainer Defaults

基于页面的说明视频默认设置

For AI/tool tutorials, page-based narrated explainers, and long-form Chinese teaching videos, apply these defaults unless the user gives a conflicting reference:

For 16:9 Xiaohei explainers, read
```
references/xiaohei-16x9-layout.md
```
before building. Copy or adapt the template asset in
```
assets/xiaohei-16x9-template/
```
when the project is a HyperFrames page-based video.
Treat the layout as a contract, not a loose suggestion. The right-side Xiaohei image belongs to the right-middle area, not the upper-right corner. The left note card belongs to the left-middle area, not the lower-left area. Captions must stay below the content zones with a clear bottom safe margin.
Select a page variant before writing HTML:
```
standard
```
,
```
long-title
```
,
```
wide-visual
```
, or
```
close
```
. Use
```
long-title
```
for titles that would produce awkward automatic breaks; use
```
wide-visual
```
for horizontal illustrations; use
```
close
```
for final action or summary pages.
Manually plan title and subtitle line breaks for every page. Do not rely on automatic wrapping when a Chinese title is long enough to create orphan characters, single-word lines, or large gaps.
Use the
```
mobile-readable
```
typography scale in
```
references/xiaohei-16x9-layout.md
```
as the default for 16:9 Xiaohei page videos. Most distribution is phone-first, and horizontal pages become small on mobile. Only use the smaller
```
desktop-compact
```
exception for explicit desktop demos, projection, course-screen playback, or similarly large-screen viewing. Do not shrink all text to make crowded pages fit; first shorten copy, split lines manually, or choose a layout variant.
Use a light white-first layout by default: white background, black text, meaningful Xiaohei illustrations, and small red/orange/blue accents only for emphasis, warnings, arrows, feedback, or state. Do not drift into dark, neon, purple, or SaaS-promo styling by default.
Treat Xiaohei illustrations as explanation, not decoration. Each page's illustration must show the core action or concept being explained, such as fact-checking, division of labor, orchestration, memory, handoff, or acceptance.
Do not let a repeated page template leave a persistent empty lower-left or lower-middle area just because the Xiaohei illustration sits on the right. For teaching videos, first consider adding useful light content there rather than simply scaling the illustration larger. Good fillers are role notes, input/action/output cards, anti-pattern reminders, acceptance criteria, priority lists, or "next time use this" prompts.
Keep these filler modules compact and information-bearing. They should help the viewer remember or apply the current page, not act as decorative labels. If the content does not add a useful judgment, leave the space intentional instead of filling it mechanically.
Reserve non-overlapping zones for narrated pages: upper content and lower captions. Captions must not cover body content.
Use
```
skill-captions
```
and its
```
anchor-dark
```
panel by default. It must wrap the text rather than span the frame, keep one fixed bottom anchor, scale from the 1080p baseline, and avoid text stroke, heavy shadow, and live blur.
Avoid moving, zooming, drifting, or shaking the entire page just to create motion. A stable page base is more important than constant camera movement.
Put motion on components: title, eyebrow, accent line, cards, bullets, Xiaohei image, and captions may enter or settle while the page base, and content frame remain fixed.
Reveal necessary components progressively with the narration. Do not show all page content at once when the voiceover is still introducing pieces. Prefer a sequence such as eyebrow -> title -> accent -> illustration -> bullets/cards.
For every narrated hero term, name, number, or reveal, derive its semantic cue from the final word timestamps and apply the semantic-cue micro-sweep in
```
references/delivery-gates.md
```
. The hero must become readable on the spoken cue and finish settling while that phrase is still active; a scene midpoint alone is not enough evidence.
Prefer reveal, wipe, fade-in, and subtle settle motion. Motion should make the explanation feel spoken into view, not make the page perform for its own sake.
Element variation is a hard requirement, not taste: adjacent pages must not present an identical set of info components, and every page must show at least one component-level change (a new element entering, an emphasis, or a state switch) within its own duration. Long videos must avoid extended PPT-like stillness.
Page density floor: a
```
standard
```
page carries at least two info components besides the title and the Xiaohei illustration (note card, bullets, cards, labels, chapter cue). If a page intentionally stays sparser, record the reason in the project note.
For substantial style, caption, or motion changes, render a 10-second sample first and get approval before rendering the full video.
For any page-based Xiaohei video, create a pre-render contact sheet from page midpoints before the final full render. Fix layout failures at this stage, not after the final MP4. The contact sheet must be checked for visual center, title breaks, note-card position, illustration position, caption weight, and content/caption clearance.
Generate subtitles only after the final narration has been concatenated. Page renderers must consume
```
media/captions/captions.json
```
directly and must refuse a final build when
```
caption-qc.json
```
is absent or failed.

对于 AI/工具教程、基于页面的解说型说明视频和中文长篇教学视频，除非用户提供相冲突的参考，否则应用以下默认设置：

对于 16:9 Xiaohei 说明视频，在构建前读取
```
references/xiaohei-16x9-layout.md
```
。当项目是基于 HyperFrames 的页面视频时，复制或改编
```
assets/xiaohei-16x9-template/
```
中的模板资产。
将该布局视为合同，而不是宽泛建议。右侧 Xiaohei 图像应位于右侧中部，而不是右上角；左侧说明卡片应位于左侧中部，而不是左下角；字幕必须位于内容区域下方，并留出清晰的底部安全边距。
编写 HTML 前先选择页面变体：
```
standard
```
、
```
long-title
```
、
```
wide-visual
```
或
```
close
```
。如果标题自动换行会出现不自然的结果，使用
```
long-title
```
；如果插画为横向构图，使用
```
wide-visual
```
；用于最终行动或总结页面时，使用
```
close
```
。
为每个页面手动规划标题和副标题的换行。当中文标题过长，自动换行可能产生孤字、单字行或大块空隙时，不得依赖自动换行。
16:9 Xiaohei 页面视频默认使用
```
references/xiaohei-16x9-layout.md
```
中的
```
mobile-readable
```
排版比例。大多数分发场景以手机为先，而横向页面在移动设备上会显得很小。只有在明确进行桌面演示、投影、课程屏幕播放或类似大屏观看时，才可使用尺寸更小的
```
desktop-compact
```
例外。不要为了容纳拥挤页面而缩小所有文字；应优先缩短文案、手动拆分行或选择布局变体。
默认采用浅色白色优先布局：白色背景、黑色文字、有意义的 Xiaohei 插画，以及仅用于强调、警告、箭头、反馈或状态的少量红/橙/蓝强调色。不要默认偏向深色、霓虹、紫色或 SaaS 宣传风格。
将 Xiaohei 插画视为解释性内容，而非装饰。每个页面的插画都必须展示正在解释的核心动作或概念，例如事实核查、职责分工、编排、记忆、交接或验收。
如果重复使用页面模板，且 Xiaohei 插画位于右侧，导致左下或中下区域长期留空，则不得保留这种状态。对于教学视频，应优先考虑在空白区域加入有用的浅色内容，而不是简单放大插画。可使用的内容包括角色说明、输入/动作/输出卡片、反模式提醒、验收标准、优先级列表或“下次使用”提示。
这些补充模块应紧凑且包含实际信息，帮助观众记忆或应用当前页面的内容，而不是充当装饰性标签。
如果补充内容无法提供有用的判断，应有意保留空间，而不是机械填充。
为带旁白的页面划分互不重叠的区域：上方用于内容，下方用于字幕。字幕不得遮挡正文内容。
默认使用
```
skill-captions
```
及其
```
anchor-dark
```
面板。字幕应随文字换行，而不是横跨整个画面；保持固定的底部锚点，以 1080p 为基准进行缩放，并避免文字描边、厚重阴影和实时模糊。
不要仅仅为了制造运动而移动、缩放、漂移或抖动整个页面。稳定的页面基底比持续的镜头运动更重要。
将动效放在组件上：标题、眉题、强调线、卡片、项目符号、Xiaohei 图像和字幕可以进入或稳定下来，而页面基底和内容框架应保持固定。
随旁白逐步展示必要组件。当旁白仍在介绍不同部分时，不要一次性显示全部页面内容。优先采用眉题 → 标题 → 强调元素 → 插画 → 项目符号/卡片的顺序。
对于每一个由旁白引出的重点术语、名称、数字或揭示节点，必须从最终词级时间戳推导语义提示，并应用
```
references/delivery-gates.md
```
中定义的语义提示微扫。核心内容必须在对应词语被读出时变得清晰可读，并在该短语仍处于活跃状态时完成稳定；仅有场景中点不能作为充分证据。
元素变化是硬性要求，而非个人偏好：相邻页面不得呈现完全相同的信息组件组合，并且每个页面在其自身时长内至少要出现一次组件级变化，例如新元素进入、某个内容获得强调或状态切换。长视频必须避免长时间类似 PPT 的静止画面。
页面信息密度下限：
```
standard
```
页面除标题和 Xiaohei 插画外，至少包含两个信息组件，例如说明卡片、项目符号、卡片、标签或章节提示。如果某个页面有意保持更简洁，应在项目说明中记录原因。
如果样式、字幕或动效将发生较大变化，先渲染 10 秒样片并获得批准，然后再渲染完整视频。
对于任何基于页面的 Xiaohei 视频，在最终完整渲染前，使用各页面中点帧生成预渲染拼图。必须在此阶段修复布局问题，而不是在最终 MP4 完成后处理。检查拼图中的视觉中心、标题换行、说明卡片位置、插画位置、字幕权重以及内容与字幕之间的间距。
仅在最终旁白拼接完成后生成字幕。页面渲染器必须直接读取
```
media/captions/captions.json
```
，并且当
```
caption-qc.json
```
缺失或未通过时拒绝执行最终构建。

Workflow

工作流

1. Intake

1. 需求接入

Capture these decisions in plain language:

goal: teach, promote, recreate, translate, caption, package, or test
source: brief, script, URL, reference video, existing footage, audio, deck, PR
platform and ratio: 16:9, 9:16, 1:1, or user-specified; when neither the user nor a handoff contract specifies one, default to Douyin horizontal 16:9
required tool: HyperFrames, Remotion, ffmpeg, or no preference
fidelity: pixel-level, visual-level, style-level, or original creative
visual direction, defaulting to the white-first Xiaohei system above
final archive slug and expected delivery path

Ask only if the missing answer changes the production path. Otherwise make a conservative assumption and proceed.

用通俗语言记录以下决策：

目标：教学、推广、复刻、翻译、加字幕、打包或测试
来源：需求简报、脚本、URL、参考视频、现有素材、音频、演示文稿或 PR
平台和比例：16:9、9:16、1:1 或用户指定值；当用户和交接合同均未指定时，默认使用抖音横屏 16:9
所需工具：HyperFrames、Remotion、ffmpeg 或无偏好
保真度：像素级、视觉级、风格级或原创创意
视觉方向，默认采用上述白色优先 Xiaohei 系统
最终归档 slug 和预期交付路径

仅当缺失答案会改变制作路径时才提问。否则采用保守假设并继续执行。

2. Route

2. 路由

Read

references/routing.md

and select the narrowest workflow. Prefer an existing specialized skill over a broad one.

Common routes:

custom multi-scene HyperFrames video ->
```
general-video
```
+
```
hyperframes
```
short unnarrated motion piece ->
```
motion-graphics
```
product or SaaS promo ->
```
product-launch-video
```
website showcase ->
```
website-to-video
```
final-audio subtitle timeline / SRT / VTT ->
```
ra-audio-to-subtitles
```
existing Pluviobyte digital human / HeyGen avatar ->
```
heygen-digital-avatar
```
designed captions on talking-head footage ->
```
embedded-captions
```
designed overlay cards on footage ->
```
graphic-overlays
```
source download ->
```
ra-video-download
```
transcription or translation ->
```
视频转写
```
final archive and QC ->
```
ai-video-qc-delivery
```

读取

references/routing.md

，选择最聚焦的工作流。优先使用现有的专用 Skill，而不是宽泛的工作流。

常见路由：

自定义多场景 HyperFrames 视频 →
```
general-video
```
+
```
hyperframes
```
短篇无旁白动态作品 →
```
motion-graphics
```
产品或 SaaS 宣传片 →
```
product-launch-video
```
网站展示 →
```
website-to-video
```
最终音频字幕时间线/SRT/VTT →
```
ra-audio-to-subtitles
```
现有 Pluviobyte 数字人/HeyGen Avatar →
```
heygen-digital-avatar
```
出镜口播素材上的设计字幕 →
```
embedded-captions
```
实拍素材上的设计叠加卡片 →
```
graphic-overlays
```
下载源视频 →
```
ra-video-download
```
转写或翻译 →
```
视频转写
```
最终归档与 QC →
```
ai-video-qc-delivery
```

3. Plan Before Build

3. 构建前规划

Create a production note in the project directory for non-trivial work. Include:

scene list and timing
asset list and source paths
page-by-page Xiaohei illustration plan, unless the route does not use pages or illustrations
page-by-page layout variant plan, including
```
standard
```
,
```
long-title
```
,
```
wide-visual
```
, or
```
close
```
, plus manual title/subtitle line breaks where needed
page-by-page content-density plan, including how persistent blank areas such as the left-lower quadrant will be used or intentionally preserved
illustration provenance: which skill will generate each image, where the generated originals live, and where the project-local copies will be saved
asset-text inventory: visible words, labels, badges, stamps, and UI text already embedded in images, screenshots, and footage; planned overlays must not repeat the same semantic label unless the repetition is intentional and documented
audio/caption plan, including the exact final media passed to
```
ra-audio-to-subtitles
```
and the canonical caption artifact directory
voice plan: exact provider, voice id, script path, and whether fallback is forbidden for the final render. The workspace default is local IndexTTS2 voice
```
pluvio-indextts2-calm-v1
```
using the canonical lossless WAV in
```
automation/config/tts-routing.json
```
; MiniMax
```
.env
```
values never override this local-video route.
platform typography plan: use
```
mobile-readable
```
by default, or document the explicit large-screen reason for
```
desktop-compact
```
; list pages that need shorter copy or manual line breaks
caption safe zones and the selected caption mode; default to
```
anchor-dark
```
from
```
skill-captions
```
, or record the reason for a frameless exception
chapter grouping and real start-time boundaries for scene planning
component-level animation plan for each page, including the order in which key components appear with the narration
semantic-cue plan for every hero term, name, number, or reveal, using final word timestamps rather than estimated scene timing
pre-render layout preview plan: exact timestamps or scene midpoints for a contact sheet before full rendering
10-second sample plan before full rendering when the visual or motion style is not yet approved
acceptance checks
known risks or unavailable source assets

For reference-video work, the timing table must come from extracted frames, not from memory.

对于非简单工作，在项目目录中创建制作说明。其中应包括：

场景列表和时序
资产列表和源路径
逐页 Xiaohei 插画规划，除非所选路由不使用页面或插画
逐页布局变体规划，包括
```
standard
```
、
```
long-title
```
、
```
wide-visual
```
或
```
close
```
，以及必要的手动标题/副标题换行
逐页内容密度规划，包括如何使用或有意保留左下象限等长期空白区域
插画来源说明：每个图像由哪个 Skill 生成、生成的原始文件存放在哪里，以及项目本地副本保存在哪里
资产文字清单：图像、截图和素材中已经嵌入的可见文字、标签、徽章、印章和 UI 文本；计划添加的叠加层不得无意中重复相同语义标签，除非这种重复是有意为之且已记录
音频/字幕规划，包括传递给
```
ra-audio-to-subtitles
```
的确切最终媒体，以及标准字幕产物目录
声音规划，包括确切提供商、声音 ID、脚本路径，以及最终渲染是否禁止回退。工作区默认使用本地 IndexTTS2 声音
```
pluvio-indextts2-calm-v1
```
以及
```
automation/config/tts-routing.json
```
中的标准无损 WAV；MiniMax
```
.env
```
值绝不能覆盖此本地视频路由
平台排版规划：默认使用
```
mobile-readable
```
，或明确记录采用
```
desktop-compact
```
的大屏原因；列出需要缩短文案或手动换行的页面
字幕安全区域和所选字幕模式；默认使用
```
skill-captions
```
的
```
anchor-dark
```
，或记录采用无边框变体的原因
章节分组和场景规划中的真实起始时间边界
每个页面的组件级动画规划，包括关键组件随旁白出现的顺序
每个重点术语、名称、数字或揭示节点的语义提示规划，使用最终词级时间戳而非估算的场景时序
预渲染布局预览规划：列出完整渲染前用于生成拼图的准确时间戳或场景中点
视觉或动效样式尚未通过批准时，完整渲染前的 10 秒样片规划
验收检查
已知风险或缺失的源资产

对于参考视频工作，时序表必须来自提取的帧，而不是凭记忆填写。

4. Build

4. 构建

Use the selected execution skill. Keep edits scoped to the project. Do not add music, captions, scenes, or visual embellishment unless the user's goal needs them or the selected workflow requires them.

Before building a page-based Xiaohei video, generate the Xiaohei images first, store them under a project asset directory such as

assets/xiaohei-illustrations/

, then make the HyperFrames/Remotion layer load those image files directly. Do not start from a CSS-drawn Xiaohei placeholder when the requested deliverable is a Xiaohei-skill illustration video.

For 16:9 HyperFrames Xiaohei explainers, start from the template asset in

assets/xiaohei-16x9-template/

or reproduce its same layout contract. Keep the template coordinates stable unless a page variant needs a documented adjustment. If you adjust them, generate a fresh midpoint contact sheet and verify the page still passes the layout gate before rendering the final MP4.

For ordinary use of the V5 perspective-grid background, copy the canonical rendered loop from

05-视频组件/视频背景/透视网格背景/

into the project's own assets and loop it continuously as the bottom layer. Verify its SHA-256 against the bundled

assets/perspective-grid-v5/

mirror. Keep playback at

1×

and retain one continuous phase across scene boundaries; do not restart the loop on every page. Fork the editable source only when ratio, palette, geometry, or speed must change. A fork is a new background version and must repeat seam, geometry, contact-sheet, and media-spec QC before becoming a recommended baseline.

For page-based explainers, keep the base page stable and animate only the necessary components. If the result feels like the whole page is floating, drifting, or shaking, revise before rendering the full video.

After the final narration is concatenated, run

ra-audio-to-subtitles

before building the final composition. Render captions from its

captions.json

; do not derive subtitle timing from

voice_manifest.json

segment durations. A missing or failed

caption-qc.json

blocks the final render.

Before authoring the caption layer, load

skill-captions

, read its registered style contract, and consume the approved

captions.json

timestamps directly. Use

anchor-dark

unless the handoff or user selects another style. Preserve its shrink-wrap, fixed bottom anchor, native 4K redraw, and no-heavy-shadow constraints, and test it over both source footage and the lightest page background. A substantial caption-style change requires a short sample before the full render.

使用选定的执行 Skill。将修改限制在项目范围内。除非用户目标需要，或所选工作流要求，否则不要添加音乐、字幕、场景或额外视觉装饰。

构建基于页面的 Xiaohei 视频前，先生成 Xiaohei 图像，并将其保存在项目资产目录中，例如

assets/xiaohei-illustrations/

，然后让 HyperFrames/Remotion 层直接加载这些图像文件。当交付物是 Xiaohei Skill 插画视频时，不得从 CSS 绘制的 Xiaohei 占位图开始。

对于 16:9 HyperFrames Xiaohei 说明视频，从

assets/xiaohei-16x9-template/

中的模板资产开始，或复现其同等布局合同。除非页面变体需要经过记录的调整，否则保持模板坐标稳定。如果进行调整，必须重新生成中点拼图，并在渲染最终 MP4 前确认页面仍通过布局检查。

日常使用 V5 透视网格背景时，将

05-视频组件/视频背景/透视网格背景/

中的标准渲染循环复制到项目自己的资产目录，并将其作为底层持续循环播放。对照内置的

assets/perspective-grid-v5/

镜像验证其 SHA-256。保持

1×

播放，并在场景边界之间延续同一相位；不得在每个页面重新启动循环。只有在比例、色板、几何结构或速度必须改变时，才分叉可编辑源。分叉后的版本属于新的背景版本，必须再次完成接缝、几何结构、拼图和媒体规格 QC，之后才能成为推荐基线。

对于基于页面的说明视频，保持页面基底稳定，仅动画化必要组件。如果结果像整个页面都在漂浮、漂移或抖动，应在渲染完整视频前修改。

最终旁白拼接完成后，在构建最终合成之前运行

ra-audio-to-subtitles

。字幕应从其

captions.json

渲染；不得根据

voice_manifest.json

的分段时长推导字幕时序。缺失或未通过的

caption-qc.json

将阻止最终渲染。

编写字幕层之前，加载

skill-captions

，读取其已注册的样式合同，并直接使用已批准的

captions.json

时间戳。除非交接文件或用户选择了其他样式，否则使用

anchor-dark

。保留其自动收缩包裹、固定底部锚点、原生 4K 重绘和无厚重阴影等约束，并在源视频素材和最浅色页面背景上进行测试。字幕样式发生重大变化时，必须先制作短样片，再进行完整渲染。

5. QC And Archive

5. QC 与归档

Read

references/delivery-gates.md

. At minimum verify:

for 待制作 queue jobs:
```
automation/scripts/check_delivery.py
```
exits 0 against the handoff file, project dir, and final MP4; its output is pasted into the handoff file's 制作端回执,
```
status
```
becomes
```
已完成
```
, and
```
01-内容生产/视频工作台/已制作/<月上旬|月下旬>/<日期-主题>/
```
is assembled with the handoff file, the final MP4 copy under
```
成片/
```
, and key QC frames under
```
质检/
```
final MP4 exists in the unified workbench delivery location:
```
视频工作台/已制作/<月上旬|月下旬>/<日期-主题>/成片/
```
for every production route
width, height, frame rate, duration, video codec, audio codec, channels
representative contact sheets or side-by-side comparisons
page/scene coverage: every page has a useful illustration or a stated reason for skipping it
for page-based 16:9 Xiaohei videos, the pre-render contact sheet passes the
```
xiaohei-16x9-layout
```
gate: right visual centered, left note card centered, no awkward title wrapping, captions below content, and no persistent bottom-heavy page
for 16:9 Xiaohei page videos, the contact sheet and representative single frames use the
```
mobile-readable
```
typography scale by default and remain readable without zooming when reviewed at reduced size; any
```
desktop-compact
```
exception must be explicitly justified in the project note
Xiaohei provenance: when Xiaohei is requested or selected by default, the project contains the generated image assets, the render references those assets directly, and a contact sheet confirms the assets appear in the final scenes
captions do not cover content; the default
```
anchor-dark
```
panel from
```
skill-captions
```
wraps the text, preserves one fixed baseline, remains inside the lower safe zone, and introduces no heavy-filter warning or full-width lower-third
caption artifacts come from
```
ra-audio-to-subtitles
```
,
```
caption-qc.json
```
reports
```
status: pass
```
, timing source is
```
volcengine-word-timestamps
```
, and alignment coverage meets the contract (0.90 by default)
component order and animation are visible in extracted frames
asset-embedded text has been compared with authored overlays/chrome and no accidental duplicate semantic label remains
every narrated hero term, name, number, or reveal passes the semantic-cue micro-sweep defined in
```
references/delivery-gates.md
```
subtitles/audio sync when captions or narration exist
voice provenance: final narration must identify local IndexTTS2, voice id
```
pluvio-indextts2-calm-v1
```
, and the canonical reference WAV SHA-256. If the manifest shows
```
say
```
, browser/generic speech, MiniMax, another cloud provider, or any fallback segment, the MP4 is a draft and must not be archived as the recommended final.
reference alignment report when a reference video exists
when the V5 perspective grid is active: the project uses the canonical loop at
```
1×
```
, the phase remains continuous across scene boundaries, and foreground content does not destroy the intentional center whitespace or reduce text readability
publish-title handoff: queue jobs skip this entirely — candidates already live in the handoff file's 标题候选 section and the user picks manually at publish time; do not create
```
视频标题.md
```
. Only for non-queue jobs, use
```
ra-video-title
```
to produce 8-12 candidates after QC, return them for user selection, then create or update
```
视频标题.md
```
next to the final MP4. Do not silently overwrite an existing title unless the user has selected the replacement.

Contact sheets and QC frames are required for review. Keep full QC in the engineering project and copy the key frames into the unified archive at

已制作/<月上旬|月下旬>/<日期-主题>/质检/

for every production route.

读取

references/delivery-gates.md

。至少验证以下内容：

对于
```
待制作
```
队列任务：针对交接文件、项目目录和最终 MP4 运行
```
automation/scripts/check_delivery.py
```
时应以状态码 0 退出；其输出粘贴到交接文件的
```
制作端回执
```
中；
```
status
```
变为
```
已完成
```
；并组装
```
01-内容生产/视频工作台/已制作/<月上旬|下旬>/<日期-主题>/
```
，其中包含交接文件、
```
成片/
```
下的最终 MP4 和
```
质检/
```
下的关键 QC 帧
每条制作路由的最终 MP4 都存在于统一工作台交付位置：
```
视频工作台/已制作/<月上旬|下旬>/<日期-主题>/成片/
```
宽度、高度、帧率、时长、视频编解码器、音频编解码器和声道数
有代表性的拼图或并排对比图
页面/场景覆盖率：每个页面都有有用的插画，或记录了跳过插画的原因
对于基于页面的 16:9 Xiaohei 视频，预渲染拼图必须通过
```
xiaohei-16x9-layout
```
检查：右侧视觉居中、左侧说明卡片居中、标题无尴尬换行、字幕位于内容下方，且页面不存在持续性的底部过重问题
对于 16:9 Xiaohei 页面视频，拼图和代表性单帧默认使用
```
mobile-readable
```
排版比例，并在缩小审阅时仍保持可读，无需放大；任何
```
desktop-compact
```
例外都必须在项目说明中明确说明理由
Xiaohei 来源验证：请求或默认选择 Xiaohei 时，项目中包含生成的图像资产，渲染直接引用这些资产，并且拼图确认资产已出现在最终场景中
字幕不得遮挡内容；
```
skill-captions
```
的默认
```
anchor-dark
```
面板应随文字换行、保持固定基线、位于底部安全区域内，并且不产生重度滤镜警告或全宽下三分之一字幕条
字幕产物来自
```
ra-audio-to-subtitles
```
，
```
caption-qc.json
```
报告
```
status: pass
```
，时间来源为
```
volcengine-word-timestamps
```
，对齐覆盖率符合合同要求（默认 0.90）
从提取帧中可以确认组件顺序和动画
已将素材内嵌文字与编写的叠加层/界面元素进行比较，不存在意外重复的语义标签
每个由旁白引出的重点术语、名称、数字或揭示节点都通过
```
references/delivery-gates.md
```
中定义的语义提示微扫
存在字幕或旁白时，字幕与音频同步
声音来源：最终旁白必须标明使用本地 IndexTTS2、声音 ID
```
pluvio-indextts2-calm-v1
```
，以及标准参考 WAV 的 SHA-256。如果清单显示使用了
```
say
```
、浏览器/通用语音、MiniMax、其他云端提供商或任何回退片段，则该 MP4 仅是草稿，不得作为推荐最终成片归档
存在参考视频时，提供参考对齐报告
启用 V5 透视网格时：项目使用
```
1×
```
的标准循环，相位在场景边界之间保持连续，前景内容不会破坏有意保留的中心留白或降低文字可读性
发布标题交接：队列任务完全跳过此步骤——候选标题已位于交接文件的
```
标题候选
```
部分，由用户在发布时手动选择；不得创建
```
视频标题.md
```
。仅对于非队列任务，在 QC 后使用
```
ra-video-title
```
生成 8 至 12 个候选标题并返回给用户选择，然后在最终 MP4 旁创建或更新
```
视频标题.md
```
。除非用户已选择替换项，否则不得静默覆盖已有标题。

拼图和 QC 帧是审阅所必需的。将完整 QC 保留在工程项目中，并为每条制作路由将关键帧复制到统一归档目录

已制作/<月上旬|月下旬>/<日期-主题>/质检/

。

Output

输出

Return the recommended final MP4 path, the key QC artifact path, confirmed specs, and any limitation that remains. For 待制作 queue jobs, also return the updated handoff file path under

已制作/

with the checker result. Keep the response concise.

返回推荐的最终 MP4 路径、关键 QC 产物路径、已确认的规格，以及仍然存在的任何限制。对于

待制作

队列任务，还需返回

已制作/

下更新后的交接文件路径和检查器结果。回复应保持简洁。