Prototype · Preview: every server in the official registry, synced every 30 minutes. Sign in with Google to write reviews.Prototype · updated every 30 min
PopNx MCPMCP servers for people who use AI · explained in plain words

video-talkcraft Skill

Repository ↗

by Vincentwei1021★ 2,4201 skills

Install

What it does

Agent skill that turns Claude Code / Codex into a motion-design studio for voiceover-driven explainer videos — word-level voiceover sync, 109 motion recipe cards, an anti-slideshow camera system, Remotion rendering.

  • Connects Instruction files the AI reads, with steps that call an outside service
  • Account or key Some steps need a key for FISH_AUDIO_API_KEY (per the README)
  • Category Design and web UI
For makers: add a badge to your README

If you made this, paste this code into your README. The badge links back to this page.

[![Featured on PopNx MCP](https://mcp.popnx.com/badge/skill/vincentwei1021-video-talkcraft.svg)](https://mcp.popnx.com/en/skills/vincentwei1021-video-talkcraft)

Use together with

MCP servers that connect the AI to services for the same work

Next AI Draw.io

draw.io diagrams your AI agent draws on a live browser canvas that you can also edit by hand

by DayuanJiangWidely used↓ 21.7K/wk★ 37.1K Stars for the whole repoUpdated 6 days ago

MCP · Connects a serviceRuns on your computer

Pascal 3D Editor

Create, inspect, validate, and save editable 3D building scenes with Pascal's MCP server.

by pascalorgWidely used↓ 1K/wk★ 25.6K Stars for the whole repoUpdated 2 days ago

MCP · Connects a serviceComputer or cloud✓ service owner · pascal.app

Semiotic

Verified React chart generation: select, validate, repair, render, and inspect charts through MCP.

by nteractWidely used↓ 8.5K/wk★ 2.7KUpdated yesterday

MCP · Connects a serviceComputer or cloud

Similar Skills

Same kind of work, ordered like the Skills list

Compared with similar ones
This one
video-talkcraft
impeccablediagram-designfrontend-slidesguizang-ppt-skill
MakerVincentwei1021pbakauscathrynlaveryzarazhangruiop7418
Skills11111
GitHub stars2,420113,04268,29632,57930,091
Last code changeNov 19, 2026Nov 19, 2026Nov 19, 2026Jun 24, 2026Oct 30, 2026
License–Apache-2.0MITMITAGPL-3.0

1 skills (instruction sets) by Vincentwei1021 that an AI reads and follows · not an MCP server

How a Skill differs from MCP

MCP = the hands

Connects the AI to an outside service or data, such as reading Notion or creating an invoice. A program or server has to be running.

Skill = the manual

A text file that teaches the AI how to do a task. It connects to nothing.

Some overlap: a few Skills run scripts that call tools, and a few MCP servers only send knowledge. Start from the task you want done.

Works with

Claude CodeCodex

From the maker’s README and the folders in the repo

Install

Is it safe?

  • Who made it

    GitHub account Vincentwei1021 (person), 10 years old

  • Use and upkeep

    ★ 2,420 stars · code last changed Nov 19, 2026

  • Any scripts?

    102 scripts that can run on a computer: runtime/check-runtime.sh, runtime/link-runtime.sh, scripts/pipeline_state.mjs, scripts/workbench_contract_lint.py and 98 more

  • Uses the internet?

    Could not check every script (some are too large); read them first

  • Changes machine-wide settings?

    No hooks or writes to machine settings found

Checked automatically from the repo files at commit 75e266f (read only, nothing was run). Not a safety guarantee: read it and decide for yourself

First: which AI do you use?

Pick your app and this page shows only how to install it there

Option 1Upload to claude.ai (web and app)

  1. Download from GitHub

    This skill has no license, so we do not hand out a ZIP from our site. Download it from GitHub and arrange the folder yourself, or use "Let the AI install it" above, which is easier

    Download from GitHub (whole repo) ↗

    Unzip it, then zip only the video-talkcraft folder again. The ZIP must have a video-talkcraft folder with SKILL.md at the top level, no loose files

  2. Open Skills on claude.ai and upload

    +Create skillUpload a skillpick video-talkcraft.zip

    Open claude.ai/customize/skills ↗

  3. No Upload button? Turn on file creation first

    SettingsCapabilitiesCode execution and file creation

    Turn on Code execution and file creation first (newer menus: Create and edit files) · personal plans: Settings → Capabilities · Team/Enterprise: an admin turns it on under Organization settings → Plugins & skills · every plan, including Free

Upload once; it works in claude.ai and Claude Desktop, and in recent Claude Code signed in with a Claude account (not with an API key)

Cowork and cloud sessions use the skills turned on in your claude.ai account, not ~/.claude/skills on your computer (Claude Desktop scheduled tasks run locally and still load it)

Checked against official docs Nov 12, 2026 · Source: support.claude.com, support.claude.com, code.claude.com

Option 1Let Claude Code install it

Copy this sentence and paste it into a Claude Code chat

Where to install
Please install the skill video-talkcraft from https://github.com/vincentwei1021/video-talkcraft (commit 75e266f) for all my projects, at ~/.claude/skills/video-talkcraft. First show me every file and tell me whether it has scripts or changes machine settings, then tell me the full path you will install to and wait for me to agree.
Please install the skill video-talkcraft from https://github.com/vincentwei1021/video-talkcraft (commit 75e266f) into the project this chat is open in (at .claude/skills/video-talkcraft). First show me every file and tell me whether it has scripts or changes machine settings, then tell me the project name and the full path you will install to and wait for me to agree.

For one project, paste it into a chat that is open in that project’s folder · The AI always shows you the path before installing

The AI tells you what it will do first and asks your permission at every step

Restart Claude Code, then type /video-talkcraft or ask for a matching task and Claude picks it up

Checked against official docs Nov 12, 2026 · Source: code.claude.com

For Terminal users

A Skill is instructions the AI will follow. Read the files before installing, and install only from makers you trust.

The maker’s recommended way

Install Node.js (if you don’t have it)

Already installed? Check first · Open Terminal, type this command and press Enter

node -v
  • A version number such as v26.13.1 = installed, skip this step
  • "command not found" (on Windows: "is not recognized") = not installed yet, or installed but Terminal needs to be closed and reopened
How to open Terminal

Mac: press ⌘ + Space, type Terminal, press Enter

Windows: open Start, type Windows PowerShell, then Open

Not installed: Open the official download page, pick the LTS version and the click-through installer: .pkg on Mac · .msi on Windows. Double-click and follow it to the end. No curl command needed

Open the Node.js download page ↗

When it’s done, close Terminal, open it again and run the same command. You should see a version number: node -v

Install once; it works for everything that needs Node.js

npx skills add Vincentwei1021/video-talkcraft

Copied from the maker’s README; it downloads and runs the maker’s code

  1. Open Terminal

    Mac: press Command+Space and type Terminal · Windows: the command below does not work in PowerShell; use “Let AI install it” above instead

  2. Copy the skill folder to ~/.claude/skills

    This downloads the repo to a temporary folder and copies only the folders that contain a SKILL.md

    git clone --depth 1 https://github.com/vincentwei1021/video-talkcraft.git /tmp/popnx-skill-vincentwei1021-video-talkcraft-75e266f && mkdir -p ~/.claude/skills && cp -R '/tmp/popnx-skill-vincentwei1021-video-talkcraft-75e266f/.' ~/.claude/skills/video-talkcraft
  3. Restart Claude Code, then type /video-talkcraft or ask for a matching task and Claude picks it up

Skills you install yourself work in the ChatGPT desktop app, Codex CLI and the Codex IDE extension · ChatGPT on the web and mobile cannot use skills you install yourself, only skills that come with a plugin from the directory

ChatGPT desktop app

The ChatGPT desktop app has Codex built in, and Codex reads skills from ~/.agents/skills

Option 1Let ChatGPT install it

Open Codex in the ChatGPT desktop app and paste this sentence into the chat

Where to install
Please install the skill video-talkcraft from https://github.com/vincentwei1021/video-talkcraft (commit 75e266f) for all my projects, at ~/.agents/skills/video-talkcraft. First show me every file and tell me whether it has scripts or changes machine settings, then tell me the full path you will install to and wait for me to agree.
Please install the skill video-talkcraft from https://github.com/vincentwei1021/video-talkcraft (commit 75e266f) into the project this chat is open in (at .agents/skills/video-talkcraft). First show me every file and tell me whether it has scripts or changes machine settings, then tell me the project name and the full path you will install to and wait for me to agree.

The AI always shows you the path before installing

The AI tells you what it will do first and asks your permission at every step

Option 2Put the folder there yourself

Put the video-talkcraft folder at ~/.agents/skills/video-talkcraft

~ is your home folder ($HOME)

For Terminal users

A Skill is instructions the AI will follow. Read the files before installing, and install only from makers you trust.

  1. Open Terminal

    Mac: press Command+Space and type Terminal · Windows: the command below does not work in PowerShell; use “Let AI install it” above instead

  2. Copy the skill folder to ~/.agents/skills

    This downloads the repo to a temporary folder and copies only the folders that contain a SKILL.md

    git clone --depth 1 https://github.com/vincentwei1021/video-talkcraft.git /tmp/popnx-skill-vincentwei1021-video-talkcraft-75e266f && mkdir -p ~/.agents/skills && cp -R '/tmp/popnx-skill-vincentwei1021-video-talkcraft-75e266f/.' ~/.agents/skills/video-talkcraft

After installing

  • ChatGPT desktop app: open Skills in the sidebar to see that the skill is there
  • To use it: in ChatGPT type @ and pick the skill · in Codex CLI or the IDE extension run /skills or type $ and the skill name
  • Codex picks up newly installed skills by itself; if one does not show up, restart Codex

Checked against official docs Nov 15, 2026 · Source: learn.chatgpt.com

ChatGPT on the web and mobile

Skills you install yourself do not work here; only skills that come with a plugin from the directory

To use this skill with ChatGPT, use the ChatGPT desktop app as above

Checked against official docs Nov 15, 2026 · Source: learn.chatgpt.com

Option 1Let Codex install it

Copy this sentence and paste it into a Codex chat · Works in Codex inside the ChatGPT desktop app, Codex CLI and the IDE extension

Where to install
Please install the skill video-talkcraft from https://github.com/vincentwei1021/video-talkcraft (commit 75e266f) for all my projects, at ~/.agents/skills/video-talkcraft. First show me every file and tell me whether it has scripts or changes machine settings, then tell me the full path you will install to and wait for me to agree.
Please install the skill video-talkcraft from https://github.com/vincentwei1021/video-talkcraft (commit 75e266f) into the project this chat is open in (at .agents/skills/video-talkcraft). First show me every file and tell me whether it has scripts or changes machine settings, then tell me the project name and the full path you will install to and wait for me to agree.

For one project, paste it into a chat that is open in that project’s folder · The AI always shows you the path before installing

The AI tells you what it will do first and asks your permission at every step

Codex picks up newly installed skills by itself; if one does not show up, restart Codex · To use it: in ChatGPT type @ and pick the skill · in Codex CLI or the IDE extension run /skills or type $ and the skill name

Checked against official docs Nov 15, 2026 · Source: learn.chatgpt.com

For Terminal users

A Skill is instructions the AI will follow. Read the files before installing, and install only from makers you trust.

  1. Open Terminal

    Mac: press Command+Space and type Terminal · Windows: the command below does not work in PowerShell; use “Let AI install it” above instead

  2. Copy the skill folder to ~/.agents/skills

    This downloads the repo to a temporary folder and copies only the folders that contain a SKILL.md

    git clone --depth 1 https://github.com/vincentwei1021/video-talkcraft.git /tmp/popnx-skill-vincentwei1021-video-talkcraft-75e266f && mkdir -p ~/.agents/skills && cp -R '/tmp/popnx-skill-vincentwei1021-video-talkcraft-75e266f/.' ~/.agents/skills/video-talkcraft
  3. Restart Codex, then ask for a matching task; Codex lists the skill

Option 1Let Cursor install it

Copy this sentence and paste it into a Cursor chat

Where to install
Please install the skill video-talkcraft from https://github.com/vincentwei1021/video-talkcraft (commit 75e266f) for all my projects, at ~/.cursor/skills/video-talkcraft. First show me every file and tell me whether it has scripts or changes machine settings, then tell me the full path you will install to and wait for me to agree.
Please install the skill video-talkcraft from https://github.com/vincentwei1021/video-talkcraft (commit 75e266f) into the project this chat is open in (at .cursor/skills/video-talkcraft). First show me every file and tell me whether it has scripts or changes machine settings, then tell me the project name and the full path you will install to and wait for me to agree.

For one project, paste it into a chat that is open in that project’s folder · The AI always shows you the path before installing

The AI tells you what it will do first and asks your permission at every step

In Cursor’s Agent chat, type / and pick video-talkcraft

Checked against official docs Nov 12, 2026 · Source: cursor.com

For Terminal users

A Skill is instructions the AI will follow. Read the files before installing, and install only from makers you trust.

  1. Open Terminal

    Mac: press Command+Space and type Terminal · Windows: the command below does not work in PowerShell; use “Let AI install it” above instead

  2. Copy the skill folder to ~/.cursor/skills

    This downloads the repo to a temporary folder and copies only the folders that contain a SKILL.md

    git clone --depth 1 https://github.com/vincentwei1021/video-talkcraft.git /tmp/popnx-skill-vincentwei1021-video-talkcraft-75e266f && mkdir -p ~/.cursor/skills && cp -R '/tmp/popnx-skill-vincentwei1021-video-talkcraft-75e266f/.' ~/.cursor/skills/video-talkcraft
  3. In Cursor’s Agent chat, type / and pick video-talkcraft

Option 1Let the AI install it

For other apps that read skills. The sentence gives the paths for Claude Code, Codex and Cursor so the AI picks the one for your app · Copy this sentence into the app’s chat

Where to install
Please install the skill video-talkcraft from https://github.com/vincentwei1021/video-talkcraft (commit 75e266f) for all my projects, at ~/.claude/skills/video-talkcraft (in Codex use ~/.agents/skills/video-talkcraft, in Cursor ~/.cursor/skills/video-talkcraft). First show me every file and tell me whether it has scripts or changes machine settings, then tell me the full path you will install to and wait for me to agree.
Please install the skill video-talkcraft from https://github.com/vincentwei1021/video-talkcraft (commit 75e266f) into the project this chat is open in (at .claude/skills/video-talkcraft; in Codex use .agents/skills/video-talkcraft, in Cursor .cursor/skills/video-talkcraft). First show me every file and tell me whether it has scripts or changes machine settings, then tell me the project name and the full path you will install to and wait for me to agree.

For one project, paste it into a chat that is open in that project’s folder · The AI always shows you the path before installing

The AI tells you what it will do first and asks your permission at every step

Checked against official docs Nov 12, 2026 · Source: code.claude.com, learn.chatgpt.com, cursor.com

For Terminal users

Other apps that support Agent Skills (such as Gemini CLI or GitHub Copilot) use different folders. Check that app’s guide, then copy the folder with the SKILL.md there.

View on GitHub ↗

Skill folders checked against the official Claude Code, Codex and Cursor guides on Nov 11, 2026

✓ Commands checked Nov 19, 2026

View the original SKILL.md

The text below comes from the maker’s repo, not from PopNx MCP

Showing the first 8,000 characters · Open the full file on GitHub

---
name: video-talkcraft
description: 终极口播视频 skill:中文口播稿 + 成品配音 → CPU 字级时间戳 → SHOTBOOK 层矩阵分镜 → Remotion 电影感成片(横屏默认/竖屏)。当用户要"做口播视频"、"解说/科普视频"、"把文案变成视频"、"给配音配画面动效"时使用。默认使用成品配音,可选 Fish Audio 从稿子合成配音与时间戳;数字人生成技术不在本 skill 内(人物素材是输入)。含统一视觉语言(Apple 范式)、108 张动效配方卡、镜头三面分层工作单、七层镜头反PPT系统(极缓推拉相机/让位,运动做减法)、六式运动承接转场(每式一卡)、长镜头世界画布、anime.js+three.js 桥、自动静止检测 + 独立 subagent 评估循环。
---

# video-talkcraft — 口播视频 skill

三大来源合体:**管线**(配音→字级时间戳→Remotion)+ **词汇**(108 张动效配方卡,全配可播 demo + 自包含 tsx)+ **镜头**(七层模型反 PPT 系统)+ **视觉语言**(Apple 范式默认版)。

核心范式:解说词驱动画面,每句都要有**活的画面响应**(相机极缓推拉/已有元素的变化),
但**新元素只在语义拍边界进场,禁止机械的"一句一个新元素"**(一句一元素是堆积型凌乱的制度根源;
分镜按语义段落切,排版预算见 cinematography.md §4.5);
**一个节拍只有一个主角,说完就让位**;字幕句边界 = 全片时间锚点。

## 流程

```
① 文案 → ② 配音输入+时间戳(本机CPU) → ③ 素材 → ④ SHOTBOOK 层矩阵 → ⑤ 实现(全局系统 → 首镜先做先确认 → 其余镜头)
                                → ⑥ 渲染 → ⑦ 三重验收(机器闸全过 + 1 轮审片修 P0/P1)→ ⑧ 交付(可选续审 ≤3 轮)
```

## ⓪ 开工体检:统一依赖 + 画幅与视觉语言
**依赖只有一份**:所有口播工程的 `remotion/node_modules` 与 `workbench/node_modules` 都是指向 `<skill根>/runtime/node_modules` 的软链,版本只在
`runtime/package.json` 里钉(`@remotion/*` 全家同号)。**新片开工第一件事**(用户 2026-09-15 定版:每次新做视频都要检查更新):
```bash
bash <skill根>/runtime/check-runtime.sh --upgrade   # 装好共享依赖 + 共享无头浏览器 + 链好 workbench;Remotion 有新版就全家升级并冒烟渲 1 帧,不过自动回滚
```
末行 `[runtime] OK` 才往下走;升了级要把 `runtime/package.json` / `package-lock.json` 的改动提交。**制作中途不升级**(母版段缓存不认依赖版本,混版本渲出的段不一致)。
工程目录里**永远不跑 `npm install`**(会写进共享目录);要加包 → 加进 `runtime/package.json`,`npm ci`,再提交。

### 画幅与视觉语言(开工先定,全流程引用)
- **画幅默认横屏 1920×1080**(用户偏好);明确要发抖音/竖屏渠道才用 1080×1920
- 视觉语言:**用户明确点了风格就按用户的来**(整套 token 替换);没点时**先从口播稿判定领域、派生本片风格档**
  (`references/design-language.md` §0:读全稿答"讲什么 / 对谁讲 / 什么口吻"→ 领域 → 风格档表给底色策略 / accent / 字体气质 / 材质与卡造型 / 图表语言 / 素材气质 / 能量档,
  写进 SHOTBOOK §0 `G0 风格档`),骨架仍是 Apple 范式(一个强调色/一个投影/底色交替分幕/两档字重/
  默认幕底:浅 `pastel-mesh-flow` / 深 `mesh-flow-dark`——12 款幕底见 §1.1,代码 `template/motion-systems/backdrop.tsx`),
  派生本片 token 落成 `theme.ts`。对任何风格都成立的只有一条:禁止逐场景随手取色。
  **卡是中性 UI,进片必须蒙皮**:动效卡的 demo / tsx 是无风格的中性呈现,复制进工程后按风格档改皮(颜色 / 字体 / 圆角 / 材质 / 图表坐标轴与标记 / 占位图形)、
  不改运动命门(时序 / 缓动 / 几何比例 / 层级),契约与反例见 design-language §0.4;SHOTBOOK 每镜写**蒙皮行**

## ① 口播稿
- 每句一个信息点,钩子在第一句(数字/冲突);13 句 ≈ 95s
- **数字一律汉字**(时间戳按文本逐字锚定,`197747` 无法与"十九万七千"的读音对位);英文品牌词直接写(中英混合对齐已验证)
- 先调研核实事实,列"事实红线清单"(不可说错的数字/未验证数据不引用)

## ② 配音输入 → 预剪 → 字级时间戳(默认本机 CPU;可选 Fish Audio)
**默认使用成品配音**:真人录音或任何 TTS 皆可,输入 = 一条完整配音(wav/mp3)+ 与之逐字一致的口播稿。
没有录音且用户选择合成时,可使用下面方案 B;它不会替换已有配音或默认的本机 CPU 流程。

**②-0 配音预剪(真人录音必做;TTS 只会压气口,可跳过)**——剪掉口水词(嗯 / 呃 / 那个…)、结巴重说、过长停顿,
**在做时间戳之前**跑(时间戳做完再动音频 = timing / beats 全体错位;预剪 → 时间戳 → 一切后续,顺序不可换):
```bash
python3 scripts/voice_trim.py audio/raw.wav script.json --dry-run --words-out audio/asr_words.json   # 先只出报告:逐条 [filler / repeat / pause] 起止 + 共剪多少;ASR 结果落盘
python3 scripts/voice_trim.py audio/raw.wav script.json --words audio/asr_words.json --out audio/full.wav \
    [--video dh/host_raw.mp4 --video-out dh/host.mp4]               # 用户认可后落盘(吃回上一步的词表,不再跑一次 ASR);同一条录音的人物视频用同一 EDL 剪(音画逐帧同长)
```
- **稿子是真值,宁漏勿错**:只剪 ASR 里**稿子没有的插入段**——全由口水词表拼成的(filler)、等于紧邻稿文的结巴 / 重说(repeat);
  ASR 听错稿子里的字永不剪(剪它就剪掉了那个字的真实发音);稿外改了措辞的重说只报告,`--cut-unmatched` 才剪。
  词表来源缺省跑 FireRed(与下面时间戳同一后端),也吃剪映等导出的**逐字** SRT(`--srt`)或词级 JSON(`--words`);
  句级 SRT 定位不到口水词,脚本会警告。无稿时只剪**整 token 文字恰为**保守词表(嗯 / 呃 / 额 / um…)的口水词——按规范化文字精确比对、不认同音(五≠唔、饿≠呃);
  紧邻重复的短词无稿只报告不剪(逐字 token 下分不清叠词与结巴)。剪同录的透明 webm 时会解一帧核 alpha,帧数对但透明丢了也算 FAIL。
- 停顿:≥0.7s 的静音压到 0.35s(留段内最安静的一窗真实房间音,不补数字零),首尾留白 0.25 / 0.5s;
  切点先找能量谷、再吸附 30fps 帧网格;`cuts.json` 是 EDL(含源 → 新时间轴映射表,剪人物视频与后续任何对账都吃它)。
- **dry-run 报告必须给用户过目再落盘**:标了「切点未落在静音里,听一下」的条目让用户听那一处;脚本只删不合成,不改语速不改音高。
  剪完人物视频后跑 ③ 的 `preflight.py --media-only` 核时长——同一 EDL 只对同一条录音成立。

字级时间戳(对**预剪后**的 `audio/full.wav`):
```bash
# 方案 A(已有录音,本机 CPU 对齐):
pip install zhconv pypinyin sherpa-onnx soundfile numpy   # 默认后端 FireRedASR2-CTC int8 的全部依赖
# 首次:下载模型 767MB(model.int8.onnx + tokens.txt)放 ~/.cache/koubo/<模型名>/,地址见脚本头注释
python3 scripts/timestamps_cpu.py audio/full.wav script.json audio/timestamps.json
# 备选(免手动下模型):pip install faster-whisper 后加 --backend whisper(首跑自动下载 460MB)
python3 scripts/make_timing.py audio/timestamps.json remotion/src/timing.json

# 方案 B(可选生成配音 + 时间戳,默认请求 s2.1-pro-free;可用性以 Fish Audio 为准):
pip install requests python-dotenv   # 还需 ffmpeg,生成前会检查
# 在 .env 配置 FISH_AUDIO_API_KEY=xxx 及可选 FISH_AUDIO_REFERENCE_ID=xxx(见 .env.example)
python3 scripts/tts_fishaudio.py script.json audio/full.wav audio/timestamps.json --timing-out remotion/src/timing.json
```
- tts_fishaudio.py:默认 `--mode sentence` 每句一次请求,插入 `--pause-sec` 指定的真实静音(默认 0.25 秒);`--mode stream` 整稿一次请求、保留自然停顿。两者均接收 UTF-8 SSE,生成结束才写文件,不提供流式播放。先解码 PCM 再拼接,偏移按采样数计算。空音频、缺失/无效对齐或文本不匹配时报错;只忽略标点、大小写与字符宽度,不自动转换数字读法。中文用接口字级锚点,英文词内插值;`match/ok` 仅确认文本映射,仍需试听。合成后若再预剪,须重新运行 CPU 对齐。
- timestamps_cpu.py:ASR 词级时间戳 → 与口播稿字符级对齐(**CJK 是可靠锚点**,
  匹配键=繁简归一+无声调拼音,同音字不算错;拉丁词各家 ASR 都常拼错,在锚点间插值)→
  每句 match 质检,<0.90 标出人工听核。默认后端 FireRedASR2-CTC(尾部最稳、零误报;整段喂入 ~200s 崩、内存平方涨,脚本默认按静音切 ≤75s 段再加回偏移,`--chunk-sec` 可调),备选 faster-whisper;
  各后端横评数据与模型下载地址见脚本头注释。
- timestamps.json schema:`{sr, total, sentences:[{i,text,start,end,match,ok,words:[{text,start,end}]}]}`
  ——words 为 CJK 逐字 + 拉丁整段 token(标点跳过);满足此 schema 的任何对齐工具都可替换。
- make_timing.py:转成 timing.json(chars 与文本逐字符 1:1,标点零时长),供 `tSay/msSay` 锚点查询
- 配音自查(耳听):无爆音/截断/误读;句间留 ~0.3s 气口,时间锚点更稳

### ②-1 语义标注(2026-09-22 起必做,正主 `references/semantic-annotation.md`)
口播稿逐句标「这句在做什么」,落成工程根 `semantics.json`——**这是选卡的需求侧产物**,没有它,选卡只剩输入类型一道过滤、剩下几十张卡靠印象挑(2026-09-21 那支片的「我是万里」就这样拿到了全库能量最低的文字卡,库里 P0 的人名条没人查到)。
```bash
python3 <skill根>/scripts/semantic_annotate.py --init     # 从 timestamps.json 出骨架(带词法提示 hint)
#   逐句填 sem(26 词封闭词表,见 taxonomy.md「语义索引」)/ weight(main 主句 = 允许进新元素、sub 陪衬句)/ entities / need
python3 <skill根>/scripts/semantic_annotate.py --stats    # 校验:与时间戳逐字一致 · 词表封闭 · 词法硬规(我是X→自我介绍、点赞订阅→号召、数量词→数据)· 每镜至少一个 main
```
标注是**派生物**:预剪 → 时间戳 → 标注 → 其后一切;改稿或重剪后 `text` 对不上就是标注失效,重新 `--init`。
**②-1 早于分镜,所以 `shot` 字段此时是空的**(SHOTBOOK 还没写):④ 写完后必须回填一次,否则下游的镜头级覆盖核不到任何东西——
```bash
python3 <skill根>/scripts/semantic_annotate.py --sync-shots   # ④ 之后:按 SHOTBOOK 标题的起止秒(或 shots.json)回填 shot,只动 shot,标注不丢
```

## ③ 素材
- **素材清单从 ②-1 的语义标注生成,不靠拍脑袋**:`need` 是 `证据` 的句去截真页 / 找出处,`身份` 要头像与账号信息,`量化` 要图表或数据源,`entities` 里的人名 / 品牌 / 地点 / URL 就是检索词。
- **先给每个镜头标素材模式(多选,可组合)**:`B-roll`(实拍视频)/ `图片`(照片 / 海报 / 插图)/ `截图`(网页 / 界面证据画面)/
  `纯动效`——如"B-roll 打底 + 截图证据卡"。新闻/信息类话题证据优先:Playwright 实时截图比泛用 B-roll 更有信息量。
  四档对应 taxonomy.md 输入类型代号 **V / 图 / 截图 / 文**,SHOTBOOK 每镜写一行 `素材:V(路径)· 图(路径)· 文`(格式见 ④)
- **实拍素材是必需项,不是可选项**(2026-09-06 硬规):全片零 B-roll / 图片 = 只有动效 + 口播人物 = 观众看到的是"讲 PPT"。
  `preflight.py` 对账 SHOTBOOK:**零 V / 图 镜头直接 FAIL**,V / 图 镜头占比 < 1/3 WARN(要在 SHOTBOOK 给依据,如证据类题材以截图为主)。
  「本片不做 B-roll」**不允许**写成设计决定;视频源搜不到就降级图片(Pexels photos 原图 → Pixabay),图片也没有才进「未完成 / 未采集清单」。
  图片与视频同源同 key,采集规格(分辨率下限、落盘目录、登记)见 `references/broll-sources.md`「配图采集」
- **真图硬规**:话题存在可截的真实页面(产品官网/GitHub/文档/画廊)时,
  成片中的浏览器/页面类镜头**禁止用代码 mock 冒充截图**——卡片 demo 里的灰条假 UI 是占位物,
  成片必须按卡片"复用指引"整块换成 `<img>` 真图;mock 只允许表现无真实对应物的示意 UI,
  且 SHOTBOOK 逐镜标注"为何无真图"。**引申**:口播讲"这样的成片/效果"时,
  示例画面必须是真成片片段(`<OffthreadVideo>` 内嵌已有成片/真机内录裁切,muted);
  讲"长页面/看板/参数页"时用 Playwright **全页长截图**(放大镜/巡航类动效直接吃真图坐标)。
  **真图上的标注坐标一律机器实测,禁目测**:页面元素用 DOM `getBoundingClientRect`、成图用逐像素量测
  (目测偏 ±30px 就会把环框到别的元素上);**会滚动/移动的真图,标注(环/框/pill)必须钉在内容坐标系上随内容动**,
  钉屏幕固定位就是错位根源
- **单视频镜头要有主题边框**(2026-09-07 用户定版):镜头里唯一主体是一段视频(录屏 / 单条 B-roll / 引用别人的成片或采访)时,
  **禁止裸贴满幅、禁止裸放白卡、也不装假播放器**(进度条 / 播放键 / 时间码一律不要——不是真播放器就是在撒谎),视频区必须包一层
  `template/components/theme-frame.tsx` 的 `<ThemeFrame kind>`:八式(复古浏览器窗口 / 杂志相框 / 35mm 胶片 / 拍立得 / 工程图纸 / 笔记本 / 邮票齿边 / 双发丝线),
  **按片子调性选一式、一片只用一式**,写进 SHOTBOOK 蒙皮行;多视频的卡(bed-echo-blur 前景 / split-60-40-story 左格 / gallery-wall-dolly)的视频区也包同一式。
  框只管造型与自己的装饰接力,整体入场 / 退场 / 极缓推归镜头层;框里的画面零处理(不滤镜 / 不缩放 / 不淡入淡出)。
  例外只有两种:视频当**底床**不当主体(bed-echo-blur / §1.2 实拍底床),以及产品界面卡(chat-gpt / claude-code 类,皮即内容)。
  静图不进框(框说"这是录像",画面不动一眼假;图用 media-pop-in / slow-push-in)。规则与八式表:design-language §1.3
- **网页拍摄不贴图**:找资料/找素材时判定可用的网页,成片里**禁止以静态截图贴屏**,
  必须像手持镜头一样"拍"它——**滚**(`evidence-scroll-tour◆`:匀速上滚 ≈10% 页高/s,讲到关

Discussion

Ask, share a tip or report that it does not work. Sorted by what people marked helpful.

No posts yet. Be the first to ask or share.

Posts here are written by users, not the PopNx team.

Sign in to post

From GitHub

GitHub stars
2,420
Maker
Vincentwei1021
License
No license stated This repo states no license. By default the author keeps all rights: ask before you share it or use it commercially.
Last code change
Nov 19, 2026
SKILL.md files in the repo (including translated copies and templates)
1
Topics
agent-skills, ai-agent, claude, claude-code, explainer-video, kinetic-typography, motion-design, motion-graphics, narration, remotion, talking-head, video-production