AI 短视频工厂

AI Short Video Factory creates MP4 videos from HTML using HyperFrames. Use for captioned talking-head edits, product launches, data visualizations, code walkthroughs, social clips, GSAP animations, transitions, audio muxing, and deterministic rendering.

AI 短视频工厂｜HTML-to-Video Studio

概述

AI 短视频工厂是一个基于 HyperFrames 的 AI 视频生成 Skill，可将文字创意、结构化脚本或内容大纲自动转化为 HTML + CSS + GSAP 动画视频，并通过 Headless Chrome + FFmpeg 确定性渲染为 MP4。无需 React，无按次收费，专为 AI Agent 设计。

核心能力：

• 🎬 从文字描述一键生成完整视频（15s ~ 3min）
• 🎨 完全自定义：布局、配色、动画、字体均由 HTML/CSS 控制
• 🔄 AI 全流程自动化：初始化 → 编写 → 动画 → 渲染 → 交付
• 🎵 支持字幕同步、原片音频保留、按需 BGM 合成与音量调校
• ✅ 自检管线确保视觉完整性和技术正确性

快速开始

一句话生成视频：

"帮我做一个 30 秒的科技资讯短视频，主题是 AI 编程工具的发展趋势"

AI 会自动完成全流程：

1. 初始化项目 → npx hyperframes init
2. 编写 HTML 视频内容（场景规划 + 动画编排）
3. 渲染为 MP4 → npx hyperframes render
4. 按需处理音频 → 保留原片音频，或在用户要求时 FFmpeg 后置混音
5. 执行自检管线 → 确保质量达标后交付

最小示例：

# 创建项目
npx hyperframes init my-video --non-interactive
# 编辑 index.html（AI 自动生成内容）
# 渲染
npx hyperframes render --non-interactive

适用场景

视频类型	典型时长	示例
科技资讯 / 行业速报	30-60s	AI 编程工具趋势、技术周报
产品发布 / 功能介绍	15-45s	新功能 demo、产品亮点
数据可视化	20-40s	市场规模、增长趋势、对比分析
教程 / 代码讲解	30-120s	技术方案讲解、代码 walkthrough
社交媒体短视频	15-30s	抖音/小红书风格竖版视频
口播 + 字幕	任意	带背景去除、字幕同步的 talking-head
品牌宣传 / 活动预热	15-45s	会议宣传、品牌介绍

---

Overview (English)

HyperFrames is an open-source (Apache 2.0) HTML-native video rendering framework by HeyGen. Write video as HTML + CSS + seekable animations, then render deterministically to MP4 via headless Chrome + FFmpeg. No React required, no per-render fees, designed for AI agents.

Prerequisites Check & Installation

Before any operation, run the environment diagnostic:

npx hyperframes doctor

This reports all dependencies and their status. If any check fails, STOP and resolve before proceeding.

Required Tools (ALL mandatory)

Tool	Minimum version	Purpose	Install command
Node.js	22+	HyperFrames CLI runtime	WorkBuddy 自动管理；或 nvm install 22 / 官网安装
FFmpeg	5.0+	Video encoding, audio muxing, frame extraction	macOS: brew install ffmpeg; Linux: apt install ffmpeg
FFprobe	(bundled with FFmpeg)	Media inspection, duration/format checks	Installed with FFmpeg
Chrome Headless Shell	Auto-managed	Frame-by-frame rendering engine	npx hyperframes browser ensure (auto-downloads)
HyperFrames CLI	0.6.90+	Composition management & rendering	npx hyperframes@latest (auto via npx)

Optional Tools (for advanced workflows)

Tool	Purpose	Install command
Docker	Containerized rendering (CI/CD, remote)	brew install --cask docker / Docker Desktop
Python 3.9+	Optional audio analysis / BGM synthesis helpers	WorkBuddy 自动管理；或系统 Python 3.9+

First-Time Setup Sequence

For a brand new environment, execute these steps in order:

# 1. Verify Node.js (should be pre-installed by WorkBuddy)
node --version  # must be >= 22.0.0

# 2. Install FFmpeg if missing
which ffmpeg || brew install ffmpeg
ffmpeg -version  # confirm installation

# 3. Download Chrome Headless Shell (auto-cached at ~/.cache/hyperframes/chrome/)
npx hyperframes browser ensure

# 4. Run full diagnostic — all checks must pass
npx hyperframes doctor

Troubleshooting Common Setup Issues

Symptom	Cause	Fix
doctor shows ✗ FFmpeg	FFmpeg not installed	brew install ffmpeg
doctor shows ✗ Chrome	First run, no cached browser	npx hyperframes browser ensure
Render hangs indefinitely	Chrome sandbox conflict in WorkBuddy	Use dangerouslyDisableSandbox: true on Bash tool
EACCES permission error	npx cache permission issue	sudo chown -R $(whoami) ~/.npm
Render produces 0-byte MP4	FFmpeg encoder issue	Check ffmpeg -encoders has libx264 and aac
npx hyperframes not found	Node/npx not in PATH	确保 Node.js 22+ 在 PATH 中；WorkBuddy 环境下使用托管 Node

Pre-Flight Check (AI Must Execute Before Every Project)

# Quick 3-command pre-flight (run at start of every video task)
npx hyperframes doctor          # full diagnostic
which ffmpeg && ffmpeg -version  # confirm FFmpeg accessible
echo "Pre-flight OK"

If doctor reports any failure, resolve it BEFORE writing any HTML. Do NOT proceed with a broken environment — it will waste render time and produce invalid output.

Workflow Decision Tree

1. 从零生成视频 → Step 1 (Init) → Step 2 (Compose) → Step 3 (Animate) → Step 4 (Lint) → Step 5 (Render)
2. 口播后期加工（加字幕/特效/换背景） → 详见「口播后期处理管线」章节完整流程
3. 已有口播 + 去背景 + 换场景 → Remove background → Write composition with layers → Add captions → Add music → Render
4. 修改已有 composition → Read files → Modify → Lint → Render
5. 数据可视化视频 → Plan data scenes → Compose with animated charts → Render
6. 多段素材拼接 → FFmpeg concat → HyperFrames 加统一字幕/特效/转场 → Render

Step 1: Initialize Project

npx hyperframes init <project-name> --non-interactive
npx hyperframes init my-video --example blank
npx hyperframes init my-video --video clip.mp4        # with existing video
npx hyperframes init my-video --audio track.mp3       # with audio
npx hyperframes init my-video --tailwind              # Tailwind v4 support

Available templates: blank, warm-grain, play-mode, swiss-grid, vignelli, decision-tree, kinetic-type, product-promo, nyt-graph

CRITICAL: Rendering Requirements

These rules are MANDATORY — violating them causes the compile phase to hang indefinitely:

1. GSAP: Keep CDN URL — Always use <script src="https://cdn.jsdelivr.net/npm/gsap@3.14.2/dist/gsap.min.js"></script>. The Compiler automatically inlines CDN scripts. NEVER use a local lib/gsap.min.js — file serving breaks during render.
2. Fonts: DO NOT use @import — Never write @import url('https://fonts.googleapis.com/...'). Instead, just declare font-family: "Inter", sans-serif in CSS. The Compiler auto-resolves common Google Fonts and caches them to ~/.cache/hyperframes/fonts/. Supported fonts include: Inter, JetBrains Mono, Roboto, etc. Noto Sans SC is NOT auto-resolved — avoid it.
3. WorkBuddy sandbox — npx hyperframes render requires dangerouslyDisableSandbox: true because it launches Headless Chrome (Puppeteer).
4. Always use --non-interactive — Required for WorkBuddy automation.

CRITICAL: Prompt Priority & Creative Freedom Protocol

Priority Rule (MOST IMPORTANT — Governs Everything Below)

用户提示词的明确意图 > 硬约束（技术安全）> 软默认（参考建议）

Two modes of operation:

提示词类型	AI 行为
详细提示词（明确指定布局、配色、风格、动画等）	严格按提示词执行，只保留硬约束，软默认全部让位。AI 的任务是实现用户的视觉创意，不是把视频拉回模板。
简单提示词（只给主题/关键词，未指定视觉细节）	AI 自由发挥创意，参考软默认值作为起点，鼓励创新布局和视觉表达。

Creative License（创意许可）： 当用户的提示词包含具体的视觉描述、风格要求、布局指示、动效要求时，视为创意许可——AI 应优先实现用户的视觉意图，所有软默认规则自动退让。AI 不应将每个视频拉回同一个标准模板。

无论哪种模式，AI 必须在最终输出前执行 §9 AI自检管线，确保技术安全。

---

Rule Classification: Hard Constraints vs Soft Defaults

🔴 硬约束（HARD CONSTRAINTS — 永远生效，即使用户提示词冲突也不可违反）

这些是技术限制，违反会导致渲染崩溃、输出错误或不可预期行为：

#	硬约束	原因
H1	禁止 Math.random()/Date.now()/new Date()/performance.now()	非确定性渲染导致帧不一致
H2	字体仅限白名单: Inter, JetBrains Mono, Roboto, sans-serif	其他字体 lint 报错或渲染失败
H3	禁止 @import url() 引入字体	Compiler 不支持，渲染卡死
H4	GSAP CDN 必须使用 jsdelivr URL	本地文件路径在渲染环境中不可用
H5	window.__timelines 必须同步注册	异步注册导致空帧
H6	内容不可超出画布边界（任何像素）	超出部分被裁切，用户看到残缺内容
H7	禁止 inline style="top:XX%" 覆盖内容定位	百分比定位跨场景不一致，导致溢出
H8	GSAP repeat 使用 Math.floor 而非 Math.ceil	ceil 可能超出 composition 时长
H9	音频必须 FFmpeg 后置合成（≥30s视频）	HyperFrames 内置音频 ~32s 截断 bug

🟢 软默认（SOFT DEFAULTS — 仅当用户提示词未指定时生效，一旦用户有明确意图即让位）

这些是经过验证的良好实践，但不是唯一正确答案：

#	软默认	默认值	用户可覆盖场景
S1	标题区位置	top: 50px	用户要全屏标题、底部标题等
S2	内容起始位置	top: 240px	用户要居中布局、沉浸式设计等
S3	卡片间距	28-36px	用户要紧凑/宽松排版
S4	内容底部边界	980px	用户明确要底部内容
S5	左右安全边距	120px	用户要全出血设计
S6	字号范围	见参考表	用户指定特定字号风格
S7	卡片宽度	320-520px	用户要大卡片/小卡片
S8	场景过渡方式	0.6s opacity fade	用户指定滑动/缩放/3D等过渡
S9	背景风格	暗色渐变 + 网格 + 模糊光球	用户指定任何其他背景
S10	粒子数量	40 个, mulberry32 seed=42	用户不要粒子或要更多
S11	骨架结构	scene-wrapper 标准骨架	用户描述了不同的布局需求
S12	配色方案	无预设（等用户指定）	用户未指定时用中性深色/浅色

---

1. Prompt Compliance Checklist

Before writing HTML, create a compact production checklist with:

• Topic/title and target audience
• Required scenes or content points
• Required visual style and forbidden styles
• Required color palette — ALWAYS follow the user's prompt specification; never substitute or override
• Required BGM style, whether it must cover the whole video, and target volume
• Duration target and final scene end time
• Any requested font/card/icon scale preferences
• Creative direction signal: 用户是否给了详细视觉指示？(YES → 严格执行，软默认让位; NO → AI自由发挥，参考软默认)

After rendering, check the output against this list. If the user asked for a specific BGM style, do not replace it with a different style unless explicitly approved. If the user specified a color palette, do not swap it for a "safer" or "more generic" scheme.

2. Color Palette Rule

User's prompt is the ONLY source of truth for color decisions. The Skill provides NO default theme colors. When the user provides a color scheme (e.g., "霓虹青绿玫红橙紫"), implement it exactly. When the prompt omits color requirements, AI may freely choose a palette that best serves the content and mood — no need to ask unless the user is unsatisfied.

3. Safe Typography Scale for 1920x1080 (Soft Default — 参考范围)

These are reference ranges for when the user hasn't specified typography preferences. If the user describes a specific visual style (e.g., "极简大字报风格", "数据密集仪表盘"), AI should adapt freely while ensuring readability.

Element	Reference range	Hard max (H6: 不可超出画布)
Cover mega title	96-120px	132px
Scene title	64-84px	92px
Subtitle	38-56px	64px
Body text	28-42px	48px
Card title	32-44px	48px
Data number	72-104px	112px
Icon/emoji	56-96px	112px
Code text	26-36px	42px

Note: Hard max 仅为防溢出，不是风格限制。如果用户明确要超大标题效果且布局容得下，AI 应实现。

4. Layout Safe Zones (Soft Default — 标准布局参考)

以下是 1920x1080 的标准布局参考。当用户提示词未指定布局时使用；当用户描述了不同的布局需求（如全屏沉浸式、非规则排列、斜切设计、居中对称等），AI 应自由设计，仅确保不违反硬约束 H6（内容不超出画布）和 H7（禁止 inline top%）。

┌──────────────────────────────────────────────┐
│ Title Safe Zone: top 50px, height ≤ 150px    │ ← 标准参考
├──────────────────────────────────────────────┤
│ Content Safe Zone: top 240px ~ bottom 980px  │ ← 标准参考
├──────────────────────────────────────────────┤
│ Bottom Safe Margin: bottom 100px             │ ← 建议保留
└──────────────────────────────────────────────┘

硬约束（始终生效）：

• H6: 所有可见内容必须在 01920px (水平) 和 01080px (垂直) 范围内
• H7: 禁止 inline style="top:XX%" — 必须使用 CSS class 或固定 px 值

软默认（未指定时使用）：

• 标题区: top: 50px, 居中, height ≤ 150px
• 内容区: top: 240px, 底部不超过 980px
• 左右边距: ≥ 120px
• 标题与内容间距: ≥ 40px

Content height budget (标准布局下的参考计算):

available_height = 980px - 240px = 740px
max_card_height = (available_height - (rows-1) × gap) / rows

Example: 3 rows, gap=24px → max per row = (740 - 48) / 3 = 230px ✓
Example: 6 cards 2×3, gap=28px → max per row = (740 - 28) / 2 = 356px ✓
Example: 4 vertical items, gap=18px → max per item = (740 - 54) / 4 = 171px ✓

If content_height > available_height (overflow risk), apply fixes:

1. Reduce card padding (28px → 20px)
2. Reduce gap (36px → 24px → 18px)
3. Reduce icon/text size within cards
4. Split into two consecutive scenes

Card layout dimension reference (soft defaults):

Layout type	Card width	Max gap	Container width
3 horizontal	360-480px	36px	90% (1728px)
4 horizontal	280-380px	30px	90%
2×3 grid	320-520px	28px	1600px
2 large horizontal	680-780px	36px	90%
Vertical list (3-4 items)	90% width	20-24px	90%
Timeline (3 items)	90% width	20px	90%

5. Font White List — HARD CONSTRAINT (H2)

Only use fonts that HyperFrames Compiler can auto-resolve:

/* ALLOWED */
font-family: "Inter", sans-serif;
font-family: "JetBrains Mono", monospace;  /* for code blocks */
font-family: "Roboto", sans-serif;

/* PROHIBITED — cause lint errors or render failures */
font-family: "PingFang SC";        /* macOS only, not bundled */
font-family: "Microsoft YaHei";    /* Windows only */
font-family: "Noto Sans SC";       /* not auto-resolved */
font-family: "Source Han Sans";    /* not auto-resolved */

Chinese text rendering: rely on sans-serif fallback (renders correctly in Chrome). The visual difference is negligible for video output.

6. Deterministic Rendering — HARD CONSTRAINT (H1, H8)

HyperFrames requires frame-level determinism for multi-pass rendering:

// ❌ PROHIBITED — non-deterministic (H1)
Math.random()
Date.now()
new Date()
performance.now()  // for positioning

// ✅ REQUIRED — use seeded PRNG
function mulberry32(seed) {
  return function() {
    seed |= 0; seed = seed + 0x6D2B79F5 | 0;
    let t = Math.imul(seed ^ seed >>> 15, 1 | seed);
    t = t + Math.imul(t ^ t >>> 7, 61 | t) ^ t;
    return ((t ^ t >>> 14) >>> 0) / 4294967296;
  }
}
const rand = mulberry32(42);  // fixed seed

Also for GSAP repeat calculations (H8):

// ❌ May overshoot composition duration
repeat: Math.ceil(duration / cycle) - 1

// ✅ Guaranteed to stay within bounds
repeat: Math.floor(duration / cycle) - 1

7. Scene Duration and Root Duration

The root composition duration must equal the last scene end time:

root_duration = max(scene.data_start + scene.data_duration)

Never leave data-duration shorter than the final scene. This causes unreliable render behavior and confusing QA results.

8. BGM Must Match Prompt and Cover Full Video

For videos longer than 30s, do not rely on short audio loops inside HyperFrames. Use this reliable workflow:

1. Generate or prepare a full-length WAV whose duration is at least root_duration + 3s.
2. Render the visual video with HyperFrames (audio data-duration can match video, but DO NOT trust the rendered audio).
3. Replace/mux final audio with FFmpeg from the WAV source:

ffmpeg -y -i rendered.mp4 -i bgm_full.wav -map 0:v:0 -map 1:a:0 \
  -c:v copy -c:a aac -b:a 192k -t <root_duration+0.02> \
  -movflags +faststart final.mp4

4. Verify the final MP4 by extracting the audio stream and checking real decoded duration and per-second RMS:

ffmpeg -y -i final.mp4 -vn -ac 1 -ar 44100 extracted_audio.wav
ffprobe -v quiet -show_entries format=duration -show_entries stream=codec_type,duration -of default=noprint_wrappers=1 final.mp4
ffmpeg -ss <root_duration-12> -t 12 -i final.mp4 -af volumedetect -f null /dev/null 2>&1 | grep -E "mean_volume|max_volume"

Bundled helper option:

python scripts/verify_audio.py final.mp4 --min-duration <root_duration> --tail-seconds 12

A successful BGM check must prove:

• Extracted audio duration >= video duration - 0.1s
• No silent seconds in the final 12 seconds (RMS > -30dB per second)
• The BGM style matches the user's requested or previously approved style
• Volume is audible but not overpowering (RMS mean between -15dB ~ -20dB)

BGM generation fallback: Only generate BGM when the user explicitly requests new music or when the source material has no audio to preserve. If numpy is unavailable, use Python standard library struct + wave + math to synthesize. A 130BPM electronic track can include: kick (4-on-floor), hi-hat (8th notes), bass (sub oscillator), pad (chord progression), lead melody, and arpeggio layers. For existing edited videos with original audio, preserve and post-mux the original audio instead of synthesizing replacement music.

Bundled helper option for generated BGM projects:

python scripts/gen_bgm.py bgm.wav --duration <root_duration+3> --bpm 110 --volume 0.25

9. AI Self-Check Pipeline (Mandatory Before Delivery — 两种模式通用)

无论用户提示词是详细还是简单，最终交付前都必须执行此管线。 此管线验证的是技术安全和视觉完整性，不是风格是否符合某个模板。

Execute ALL checks in this exact order. Do NOT skip any step.

Phase A: Pre-Write Validation (before writing HTML)

#	Check	详细提示词模式	简单提示词模式
A1	Scene count × avg duration = total duration?	按提示词场景规划	AI 自行规划
A2	Content density reasonable for scene duration?	按提示词内容量	参考中文密度表
A3	Colors match user specification?	严格匹配提示词配色	AI 自由选择
A4	Font family only uses whitelist? (H2)	始终检查	始终检查
A5	是否存在 Hard Constraint 冲突?	始终检查	始终检查

Phase B: Post-Write / Pre-Render Validation

# B1: Lint must pass with 0 errors (HARD — always)
npx hyperframes lint

# B2: Check for prohibited patterns in HTML (HARD — H1, H2)
grep -n "Math.random\|Date.now\|PingFang\|Microsoft YaHei\|Noto Sans" index.html
# → must return empty

# B3: Check for inline top% overrides (HARD — H7)
grep -n 'style=.*top:.*%' index.html
# → If found: REMOVE immediately (use CSS class or fixed px)

# B4: Check root data-duration matches expected
grep 'data-composition-id.*data-duration\|data-start.*data-duration' index.html
# → Verify last scene end = root duration

Phase C: Post-Render Validation

#	Check	Command	Pass criteria
C1	Video duration	ffprobe -show_format	≥ target - 0.1s
C2	Resolution	ffprobe -show_streams	匹配用户要求的分辨率
C3	Frame rate	ffprobe -show_streams	30fps (除非用户要求 60fps)
C4	Audio duration after mux	Extract WAV, check length	≥ video duration
C5	Last 12s no silence	Per-second RMS check	All seconds > -30dB
C6	Audio mean volume	volumedetect	-15dB ~ -20dB

Phase D: Layout & Visual Integrity Check (关键步骤)

For every video, extract key frames and verify visual integrity:

# Extract key frames at scene midpoints
ffmpeg -ss <mid_time> -i final.mp4 -frames:v 1 -q:v 2 check_scene_N.jpg

通用视觉检查（两种模式都执行）：

• ✅ 所有文字和内容在画布范围内（H6）— 无裁切
• ✅ 无文字互相重叠导致不可读
• ✅ 卡片/元素间距均匀，无挤压变形
• ✅ 关键信息可读（字号足够，对比度足够）
• ✅ 动画过渡流畅，无跳切（除非用户要求跳切风格）

额外检查（仅标准布局模式——使用软默认时）：

• 标题在顶部可见，不与内容重叠
• 所有卡片在安全区范围内

If ANY check in Phases B-D fails → fix → re-render → re-verify. Maximum 2 retry cycles; if still failing after 2 retries, report the specific issue to the user.

10. Auto-Completion Defaults (Soft — 仅简单提示词模式)

当用户的提示词未指定以下细节时，AI 可参考这些经过验证的默认值作为起点，但鼓励在此基础上发挥创意：

Missing item	Default reference	AI 可自由替换？
Layout strategy	Title top:50px + Content top:240px	✅ 可用任何不违反 H6/H7 的布局
Font	Inter, sans-serif	❌ 硬约束 H2，只能用白名单字体
Audio strategy	FFmpeg post-mux from full-length WAV	❌ 硬约束 H9（≥30s 视频）
Content container	.content-area / .content-col	✅ 可用任何语义化 CSS 结构
Card gap	28-36px (horizontal), 20-24px (vertical)	✅
Particle generation	40 particles, mulberry32 seed=42	✅ 数量自由，但 PRNG 必须用 mulberry32 (H1)
Background	Dark gradient + grid + blur orb	✅ 完全自由
Transition style	0.6s opacity fade	✅ 可用滑动、缩放、模糊等
Grid background animation	20s infinite translate loop	✅
Verification pipeline	Full Phase A-D self-check	❌ 始终必须执行

Key principle: 默认值是灵感参考，不是束缚。AI 应根据视频主题和内容自然选择最佳表达方式。

11. Common Pitfalls Registry (Quick Reference)

Pitfall	Root cause	Type	Prevention
Content overflows bottom	top:XX% inline style	🔴 H7	禁止 inline top%，用 CSS class 固定 px
Audio cuts at ~32s	HyperFrames built-in audio bug	🔴 H9	Always FFmpeg post-mux
Non-deterministic frames	Math.random()	🔴 H1	Replace with mulberry32 PRNG
Lint error: unresolved font	Using PingFang SC etc.	🔴 H2	Font whitelist only
GSAP overruns composition	Math.ceil for repeat count	🔴 H8	Use Math.floor instead
Empty/black frames	window.__timelines not registered	🔴 H5	Ensure synchronous registration
Cards overlap title	transform:translate(-50%,-50%) centering	⚠️ 自检	Phase D frame spot-check 发现即修
2×3 grid uneven spacing	Card width too small for container	⚠️ 自检	Phase D 视觉检查
Scene jump cuts	Missing exit transitions	⚠️ 自检	除非用户要求跳切风格

Step 2: Write HTML Composition

A composition is an HTML file. The root container defines the video canvas.

Simple Single-Scene Example

<!doctype html>
<html>
<head>
  <script src="https://cdn.jsdelivr.net/npm/gsap@3.14.2/dist/gsap.min.js"></script>
</head>
<body>
  <div data-composition-id="main" data-start="0" data-width="1920" data-height="1080">

    <!-- Video clip: track 0, starts at 0s, plays for 10s -->
    <video id="bg-video" data-start="0" data-duration="10" data-track-index="0"
           src="background.mp4" muted playsinline></video>

    <!-- Title overlay: track 1, appears at 1s for 4s -->
    <h1 id="title" class="clip" data-start="1" data-duration="4" data-track-index="1">
      Product Launch
    </h1>

    <!-- Audio: track 2, from 0s for 10s, volume 50% -->
    <audio data-start="0" data-duration="10" data-track-index="2"
           data-volume="0.5" src="music.wav"></audio>
  </div>

  <style>
    body { margin: 0; overflow: hidden; }
    [data-composition-id="main"] {
      width: 1920px; height: 1080px;
      position: relative; background: #000;
    }
    #title {
      position: absolute; top: 50%; left: 50%;
      transform: translate(-50%, -50%);
      font-size: 96px; color: white; font-family: sans-serif;
    }
  </style>

  <script>
    window.__timelines = window.__timelines || {};
    const tl = gsap.timeline({ paused: true });
    tl.from("#title", { opacity: 0, y: 60, duration: 0.8, ease: "power3.out" }, 1);
    tl.to("#title", { opacity: 0, y: -40, duration: 0.5, ease: "power2.in" }, 4);
    window.__timelines["main"] = tl;
  </script>
</body>
</html>

Multi-Scene Standard Skeleton (Soft Default — Reference Starting Point)

For videos with 3+ scenes when the user hasn't specified a custom layout, this skeleton provides a proven starting point. If the user's prompt describes a different visual structure (e.g., full-screen transitions, split-screen, non-linear navigation, cinematic parallax), AI should design freely — only the Hard Constraints (H1-H9) remain in effect.

<!doctype html>
<html>
<head>
  <script src="https://cdn.jsdelivr.net/npm/gsap@3.14.2/dist/gsap.min.js"></script>
</head>
<body>
<div data-composition-id="main" data-start="0" data-duration="TOTAL_SECONDS"
     data-width="1920" data-height="1080">

  <!-- ===== BACKGROUND LAYER (persists entire video) ===== -->
  <div id="bg-layer" style="position:absolute;inset:0;z-index:0;">
    <!-- Grid background, particles, orbs — shared across scenes -->
    <div id="grid-bg"></div>
    <div id="particles-container"></div>
  </div>

  <!-- ===== SCENE 1: Opening ===== -->
  <div id="scene-1" class="scene-wrapper" data-start="0" data-duration="4" data-track-index="1">
    <div class="scene-title" id="s1-title">
      <!-- Main title content -->
    </div>
    <div class="content-area" id="s1-content">
      <!-- Scene content: cards, text, data -->
    </div>
  </div>

  <!-- ===== SCENE 2 ===== -->
  <div id="scene-2" class="scene-wrapper" data-start="4" data-duration="4" data-track-index="1">
    <div class="scene-title" id="s2-title">
      <h2>场景标题</h2>
      <p class="subtitle">副标题说明</p>
    </div>
    <div class="content-area" id="s2-cards">
      <!-- Horizontal card layout -->
    </div>
  </div>

  <!-- ===== SCENE N: (repeat pattern) ===== -->
  <!-- ... -->

  <!-- ===== AUDIO ===== -->
  <audio data-start="0" data-duration="TOTAL_SECONDS" data-track-index="2"
         data-volume="0.5" src="bgm.wav"></audio>
</div>

<style>
  * { box-sizing: border-box; margin: 0; padding: 0; }
  body { margin: 0; overflow: hidden; background: #0a0a0f; font-family: "Inter", sans-serif; color: #fff; }
  [data-composition-id="main"] { width: 1920px; height: 1080px; position: relative; overflow: hidden; }

  /* ===== SCENE WRAPPER ===== */
  .scene-wrapper { position: absolute; inset: 0; opacity: 0; }

  /* ===== TITLE ZONE: fixed top 50px ===== */
  .scene-title {
    position: absolute;
    top: 50px;
    left: 0;
    width: 100%;
    text-align: center;
    z-index: 10;
  }
  .scene-title h2 { font-size: 76px; font-weight: 700; margin: 0; }
  .scene-title .subtitle { font-size: 38px; opacity: 0.7; margin-top: 8px; }

  /* ===== CONTENT ZONE: fixed top 240px ===== */
  .content-area {
    position: absolute;
    top: 240px;
    left: 50%;
    transform: translateX(-50%);
    width: 90%;
    display: flex;
    justify-content: center;
    align-items: flex-start;
    gap: 36px;
    flex-wrap: wrap;
  }
  .content-col {
    position: absolute;
    top: 240px;
    left: 50%;
    transform: translateX(-50%);
    width: 90%;
    display: flex;
    flex-direction: column;
    align-items: center;
    gap: 24px;
  }

  /* ===== CARD STYLES ===== */
  .glass-card {
    background: rgba(255,255,255,0.04);
    border: 1px solid rgba(255,255,255,0.08);
    border-radius: 16px;
    padding: 28px 24px;
    backdrop-filter: blur(12px);
  }
</style>

<script>
  // ===== Seeded PRNG (mulberry32) =====
  function mulberry32(seed) {
    return function() {
      seed |= 0; seed = seed + 0x6D2B79F5 | 0;
      let t = Math.imul(seed ^ seed >>> 15, 1 | seed);
      t = t + Math.imul(t ^ t >>> 7, 61 | t) ^ t;
      return ((t ^ t >>> 14) >>> 0) / 4294967296;
    }
  }
  const rand = mulberry32(42);

  // ===== GSAP Timeline =====
  window.__timelines = window.__timelines || {};
  const tl = gsap.timeline({ paused: true });
  const TOTAL = TOTAL_SECONDS;

  // --- Scene 1: 0-4s ---
  tl.to("#scene-1", { opacity: 1, duration: 0.01 }, 0);
  tl.from("#s1-title", { scale: 0.5, opacity: 0, duration: 0.8, ease: "back.out(1.4)" }, 0.1);
  tl.from("#s1-content", { y: 40, opacity: 0, duration: 0.6 }, 0.5);
  tl.to("#scene-1", { opacity: 0, duration: 0.5 }, 3.4);  // exit at scene_end - 0.6

  // --- Scene 2: 4-8s ---
  tl.to("#scene-2", { opacity: 1, duration: 0.01 }, 4);
  tl.from("#s2-title", { x: -60, opacity: 0, duration: 0.6, ease: "power2.out" }, 4.1);
  tl.from("#s2-cards .glass-card", { scale: 0.6, opacity: 0, duration: 0.5, stagger: 0.15 }, 4.4);
  tl.to("#scene-2", { opacity: 0, duration: 0.5 }, 7.4);

  // --- Scene N: (repeat pattern) ---
  // ...

  window.__timelines["main"] = tl;
</script>
</body>
</html>

Skeleton naming conventions (recommended for consistency, not mandatory):

• Scene wrappers: #scene-1, #scene-2, ... #scene-N
• Scene titles: #s1-title, #s2-title, ... #sN-title
• Content containers: #s1-content, #s2-cards, #s3-items, ... (descriptive suffix)
• Card elements: .glass-card, .data-card, .app-card (semantic naming)
• Background: #bg-layer, #grid-bg, #particles-container

Scene wrapper pattern (recommended, AI may use alternative transition approaches if user requests):

// Scene entry (instant opacity on)
tl.to("#scene-N", { opacity: 1, duration: 0.01 }, SCENE_START);
// Scene content animations
tl.from("#sN-title", { /* entrance */ }, SCENE_START + 0.1);
tl.from("#sN-content ...", { /* entrance */ }, SCENE_START + 0.3);
// Scene exit (0.6s before end) — EXCEPT final scene
tl.to("#scene-N", { opacity: 0, duration: 0.5 }, SCENE_END - 0.6);

Data Attributes Reference

Attribute	Required	Purpose
data-composition-id	Yes	Unique ID for the composition
data-start	Yes	Start time in seconds (or clip ID ref: "el-1 + 2")
data-duration	Yes for img/div	Duration in seconds (video/audio auto-detect)
data-track-index	Yes	Track layer (same-track clips cannot overlap)
data-width / data-height	Yes (root)	Canvas dimensions (1920x1080 or 1080x1920)
data-volume	No	Audio volume 0-1 (default 1)
data-media-start	No	Trim offset into source media
data-composition-src	No	Path to external sub-composition HTML

Video and Audio Rules

• Video MUST be muted playsinline — audio is always a separate <audio> element
• Never call video.play()/audio.play() — the framework owns playback
• Never nest video inside a timed div — use a non-timed wrapper
• Long BGM warning: HyperFrames' built-in audio processing may truncate long background music around ~32s even when the source audio and data-duration are longer. Always verify the rendered MP4 by extracting its audio stream and checking the decoded audio duration/RMS; volumedetect -ss 30 -t 12 can be misleading because it only analyzes available samples.
• Reliable long-BGM workflow: render the visual video first, then replace/mux the audio with FFmpeg from a full-length WAV source: ffmpeg -i rendered.mp4 -i bgm_full.wav -map 0:v:0 -map 1:a:0 -c:v copy -c:a aac -b:a 192k -t <composition_duration> -movflags +faststart final.mp4. Verify by extracting final.mp4 audio to WAV and checking per-second RMS through the end.

Step 3: Animate with GSAP

Non-Negotiable Rules

1. All timelines MUST start { paused: true } — player controls playback
2. Register every timeline: window.__timelines["<composition-id>"] = tl
3. Duration comes from data-duration, NOT from GSAP timeline length
4. No repeat: -1 — calculate exact repeats: repeat: Math.ceil(duration / cycleDuration) - 1
5. No Math.random(), Date.now() — use seeded PRNG if needed
6. No async timeline construction — no setTimeout, await, Promises
7. Only animate visual properties: opacity, x, y, scale, rotation, color, backgroundColor
8. Never animate visibility, display
9. Offset first animation 0.1–0.3s from t=0

Scene Transitions (multi-scene compositions)

1. ALWAYS use transitions between scenes — no jump cuts
2. ALWAYS use entrance animations (gsap.from()) on every element
3. NEVER use exit animations except on the final scene — the transition IS the exit
4. Final scene only may fade elements out

Layout Before Animation

Build the end-state first as static CSS, then add motion:

1. Position elements at their most visible moment
2. Add entrances with gsap.from() — animate FROM offscreen TO CSS position
3. Add exits with gsap.to() — only on final scene

Step 4: Lint & Inspect

npx hyperframes lint              # Structure/code check (fast)
npx hyperframes lint --json       # Machine-readable
npx hyperframes inspect           # Visual layout check (launches Chrome)
npx hyperframes inspect --json    # Agent-readable findings

Fix all errors before rendering. Warnings should be addressed.

Step 5: Render

npx hyperframes render                          # Standard MP4
npx hyperframes render --quality draft          # Fast iteration (~3x faster)
npx hyperframes render --quality high --fps 60  # Final delivery
npx hyperframes render --output final.mp4       # Custom filename
npx hyperframes render --format webm            # Transparent WebM
npx hyperframes render --docker                 # Byte-level consistency

Flag	Options	Default	Notes
--output	path	renders/name_timestamp.mp4	Output path
--fps	24, 30, 60	30	60fps doubles render time
--quality	draft, standard, high	standard	draft for iteration
--format	mp4, webm	mp4	WebM supports transparency
--workers	1-8 or auto	auto	Each worker spawns a Chrome
--docker	flag	off	Reproducible output
--variables	JSON	—	Override composition variables

Deliver the rendered MP4 to user via deliver_attachments.

Media Processing

Text-to-Speech (TTS)

Local inference with Kokoro-82M. No API key needed.

npx hyperframes tts "Your script here" --voice af_heart --output narration.wav
npx hyperframes tts script.txt --voice bf_emma --output narration.wav
npx hyperframes tts --list                      # List all 54 voices

Voice selection:

Content Type	Voice	Reason
Product demo	af_heart / af_nova	Warm, professional
Tutorial	am_adam / bf_emma	Neutral, easy to follow
Marketing	af_sky / am_michael	Energetic or authoritative
Casual/social	af_heart / af_sky	Friendly, natural

Chinese voices use z prefix: e.g., zf_xiaobei, zm_yunjian.

Requirements: Python 3.8+, pip install kokoro-onnx soundfile. Non-English needs espeak-ng.

Transcription (for captions)

Local Whisper inference. No API key needed.

npx hyperframes transcribe audio.mp3                      # → transcript.json
npx hyperframes transcribe video.mp4 --model small        # Default model
npx hyperframes transcribe video.mp4 --model medium       # Better accuracy
npx hyperframes transcribe subtitles.srt                  # Import existing

Output: transcript.json with word-level timestamps:

[{"id": "w0", "text": "Hello", "start": 0.0, "end": 0.5}, ...]

CRITICAL: Never use .en models unless audio is confirmed English. .en models TRANSLATE instead of transcribe.

Background Removal

Local u2net_human_seg model. No API key.

npx hyperframes remove-background talking-head.mp4 -o transparent.webm
npx hyperframes remove-background subject.mp4 -o subject.webm --background-output plate.webm
npx hyperframes remove-background portrait.jpg -o cutout.png

Output formats:

• .webm (VP9 + alpha) — for direct use in <video> composition
• .mov (ProRes 4444) — for editing in external tools
• .png — single image cutout

Common Video Recipes

Recipe 1: Talking-Head + Captions + Background Music

# 1. Remove background from talking-head video
npx hyperframes remove-background talking-head.mp4 -o transparent.webm

# 2. Transcribe for captions
npx hyperframes transcribe talking-head.mp4 --model small

# 3. Init project and compose
npx hyperframes init captioned-video --non-interactive

Composition structure:

• Track 0: Background (gradient, image, or video)
• Track 1: Transparent talking-head (transparent.webm)
• Track 2: Animated captions (driven by transcript.json)
• Track 3: Background music (<audio> with data-volume="0.3")

Recipe 2: Product Launch Video

npx hyperframes init product-launch --example product-promo --non-interactive

Typical structure: 3-5 scenes with title → features → demo → CTA.

Recipe 3: Code Walkthrough

Structure: Terminal/editor mockup background + animated code highlights + voiceover captions.

Recipe 4: Data Visualization

Structure: Animated charts (CSS/GSAP driven), counters, stat reveals with staggered entrances.

Recipe 5: Social Media Short (Vertical 1080x1920)

Set data-width="1080" data-height="1920" on root. Refer to "Vertical Video Safe Zones (P1)" section for complete layout rules, typography scale, and safe margins. Key points: title at top 160px, content zone 400-1600px, bottom 320px reserved for platform UI, fast pacing (2-3s per scene), bold colors, single-column card layout preferred.

---

口播后期处理管线（Post-Production Pipeline for Talking-Head Videos）

本章节覆盖 对已有视频素材进行后期处理 的完整流程——加字幕、叠特效、换背景、混音 BGM、画中画等。

适用场景

场景	输入	输出
口播 + 字幕	一段录好的口播 MP4	带字幕动效的成品 MP4
口播 + 字幕 + 特效	口播 MP4	带字幕 + 粒子/光效叠加的成品
口播 + 换背景	口播 MP4（纯色/杂背景）	去背景 + 新背景的成品
多段素材拼接	多个 MP4 片段	合并 + 转场 + 统一字幕的成品
画中画	主视频 + 辅助画面	PiP 布局的成品

关键经验：口播视频修改防错清单（必须执行）

以下经验来自一次 86 秒口播视频连续迭代中反复出现的问题。处理已有口播视频时必须优先执行这些规则，避免重复踩坑。

0. 触发与能力边界：精细口播后期必须启用本 Skill

1. 用户要求“更精准时间对齐 / 更炫字幕动效 / 转场特效 / 完整口播后期处理管线”时，必须立即启用 AI 短视频工厂 Skill。 不得只用纯 FFmpeg + ASS 字幕做简单烧录后交付，因为那只能完成基础字幕，不等于完整口播后期。
2. 不得在未启用 HyperFrames 的情况下承诺“精准匹配字幕位置和特效”。 如果只是 FFmpeg 静态字幕，必须明确说明能力有限；若用户明确要求完整效果，进入 HyperFrames 管线。
3. 完整口播后期标准管线必须包含： 素材预检 → 音频提取 → Whisper/whisper.cpp 转写或时间戳分析 → 用户原文校对 → HTML Composition 编排 → GSAP 字幕/特效/转场 → HyperFrames lint → HyperFrames render → FFmpeg 后置音频合成 → 交付前验证。
4. 第一次交付就应采用正确管线。 不要先交一个“简单字幕版”再等用户指出“没有启用 HyperFrames”。如果用户一开始已经提出“自动匹配位置、加字幕和特效”，默认就是 HyperFrames 任务。
5. 若先前已经用非 HyperFrames 方案做错，必须在复盘中记录为：触发识别失败，而不是单纯“效果不好”。 后续遇到类似请求时，先加载本 Skill，再执行完整管线。

A. 源素材与音频：不要凭听感或渲染结果臆断

1. 始终先确认用户指定的原视频就是唯一音频真源。 用户说“原视频里有背景音乐”时，不得擅自判断为没有 BGM，也不得自行合成替代音乐。
2. 必须用 FFmpeg 验证尾段音频是否存在。 对用户指出的时间段（例如 1:16 后）执行：

ffmpeg -ss 76 -t 10 -i input.mov -af volumedetect -f null /dev/null 2>&1 | grep -E "mean_volume|max_volume"

如果 mean_volume/max_volume 有有效值，说明该段确实有声音，后续成品必须保留。

3. 已有原片音频时，采用“视觉渲染 + FFmpeg 后置合成原音频”的可靠流程。 不要依赖 HyperFrames 从 .mov 或长音频中直接 mux 完整音频；不要在 <audio> 中直接引用 .mov 作为音频源来保留长尾 BGM。正确做法：

# 1. 从原视频提取完整音频
ffmpeg -y -i input.mov -vn -acodec pcm_s16le -ar 44100 -ac 1 original_audio.wav

# 2. HyperFrames 渲染视觉版（可以不放 <audio>，允许 silent visual output）
npx hyperframes render --output visual.mp4

# 3. 后置合成完整原音频
ffmpeg -y -i visual.mp4 -i original_audio.wav \
  -map 0:v:0 -map 1:a:0 -c:v copy -c:a aac -b:a 192k \
  -t <video_duration> -movflags +faststart final.mp4

4. 最终必须验证音频完整性。 不仅检查 ffprobe 时长，还要检查用户指出的尾段：

ffprobe -v quiet -show_entries format=duration -show_entries stream=codec_type,duration -of default=noprint_wrappers=1 final.mp4
ffmpeg -ss 76 -t 10 -i final.mp4 -af volumedetect -f null /dev/null 2>&1 | grep -E "mean_volume|max_volume"

B. 字幕与文案：用户给的精确文案优先级最高

1. 用户明确指定某段字幕时，必须逐字使用用户原文。 不要根据 ASR、记忆或模型理解自行“纠错”产品名。例如用户指定“Codex自动化剪辑视频”，不得改成 “QDesk”“QClaw” 或其它更合理的词。
2. 每次修改字幕后必须定位对应 cap-*，只改目标时间段，不顺手改其它字幕。 对 3s-7s 这类精确时间段，先在 index.html 中找到 data-start/data-duration 覆盖该区间的字幕节点，再替换文本。
3. 避免交付前只报“已改”。 对字幕修正必须在最终回答中列出改动后的准确文本，便于用户核对。

C. 视觉迭代：严格按用户约束，不额外加效果

1. 用户要求“不要增强光效/不要粒子/不要某类动效”时，必须删除对应 CSS、DOM、GSAP tween。 不要仅设为透明或保留隐藏元素，避免后续误启用或 lint 干扰。
2. 用户要求位置变化时，按方向显式调整坐标。 例如“整体往左一点”应从 right: ... 改为明确 left: ... 或减少 right 值，并在交付说明中标注实际坐标变化。
3. 用户要求字体颜色不要为白色时，检查所有相关文本层。 数据图表要同时检查 value、label、legend，不只改一个元素。

D. 交付门禁：未验证关键问题不得交付

在每次渲染交付前，至少完成以下验证：

• 字幕关键片段：目标 cap-* 文本与用户要求完全一致。
• 音频关键片段：用户指出的尾段存在有效音量。
• 成品时长：视频流和音频流时长接近，误差 ≤ 0.1s。
• 用户本轮要求：逐项勾选，不遗漏“颜色/位置/去除效果/音频”等小项。

完整 Step-by-Step 流程

┌─────────────────────────────────────────────────────────────┐
│ Phase 1: 素材预处理                                          │
│   ① 检查素材格式 → ② FFmpeg 标准化 → ③ 提取音频             │
├─────────────────────────────────────────────────────────────┤
│ Phase 2: 分析与转写                                          │
│   ④ 语音转写 → ⑤ 字幕分组 → ⑥ 时间轴确认                   │
├─────────────────────────────────────────────────────────────┤
│ Phase 3: 合成编排                                            │
│   ⑦ 初始化项目 → ⑧ 编写 Composition HTML → ⑨ 渲染           │
├─────────────────────────────────────────────────────────────┤
│ Phase 4: 后期混音                                            │
│   ⑩ BGM 合成 → ⑪ 音量平衡 → ⑫ 交付                        │
└─────────────────────────────────────────────────────────────┘

Phase 1: 素材预处理

素材预检清单

# 检查视频信息
ffprobe -v quiet -print_format json -show_format -show_streams input.mp4

# 确认关键参数
# - 分辨率：1920x1080 或 1080x1920（竖版）
# - 帧率：25/30fps
# - 编码：H.264/H.265
# - 音频：AAC，采样率 44100/48000

常见预处理操作

# 分辨率不是 1080p → 缩放
ffmpeg -i input.mp4 -vf "scale=1920:1080:force_original_aspect_ratio=decrease,pad=1920:1080:(ow-iw)/2:(oh-ih)/2" -c:a copy normalized.mp4

# 竖版视频标准化
ffmpeg -i input.mp4 -vf "scale=1080:1920:force_original_aspect_ratio=decrease,pad=1080:1920:(ow-iw)/2:(oh-ih)/2" -c:a copy normalized_v.mp4

# 帧率标准化为 30fps
ffmpeg -i input.mp4 -r 30 -c:a copy fps30.mp4

# 提取纯音频（用于转写和混音）
ffmpeg -i input.mp4 -vn -acodec pcm_s16le -ar 16000 -ac 1 audio_for_transcribe.wav
ffmpeg -i input.mp4 -vn -acodec pcm_s16le -ar 44100 -ac 2 audio_original.wav

# 获取视频时长（秒）
ffprobe -v error -show_entries format=duration -of csv=p=0 input.mp4

多段素材拼接预处理

# 方式 1: FFmpeg concat demuxer（推荐，无重编码）
# 创建 filelist.txt:
# file 'clip1.mp4'
# file 'clip2.mp4'
# file 'clip3.mp4'
ffmpeg -f concat -safe 0 -i filelist.txt -c copy merged.mp4

# 方式 2: 需要重编码（分辨率/编码不同时）
ffmpeg -f concat -safe 0 -i filelist.txt -vf "scale=1920:1080" -c:v libx264 -c:a aac merged.mp4

# 方式 3: 在 HyperFrames 中用多个 <video> 元素分段播放（带转场）
# → 见后续 Composition 模板

Phase 2: 语音转写与字幕生成

# 转写中文口播（推荐 medium 模型，中文识别更准）
npx hyperframes transcribe input.mp4 --model medium

# 如果已有 SRT/VTT 字幕文件
npx hyperframes transcribe existing.srt

中文字幕分组规则：

规则	说明
每行最大字数	14-16 个中文字符
每组最大行数	2 行
按语义断句	在标点符号（，。！？）处断开
最短停留时间	≥ 1.2 秒（给观众阅读时间）
最长停留时间	≤ 5 秒（避免字幕"粘"太久）

Phase 3: Composition 编排

核心 HTML 模板：口播 + 字幕 + 特效

<!DOCTYPE html>
<html>
<head>
  <style>
    * { margin: 0; padding: 0; box-sizing: border-box; }

    /* === 字幕层样式 === */
    .caption {
      position: absolute;
      bottom: 100px;
      left: 50%;
      transform: translateX(-50%);
      font-family: "Inter", sans-serif;
      font-size: 42px;
      font-weight: 700;
      color: #ffffff;
      text-align: center;
      max-width: 75%;
      padding: 12px 24px;
      border-radius: 8px;
      background: rgba(0, 0, 0, 0.6);
      text-shadow: 0 2px 4px rgba(0, 0, 0, 0.5);
      /* 中文适配 */
      line-height: 1.5;
      letter-spacing: 0.02em;
    }

    /* === 特效层样式 === */
    .effect-layer {
      position: absolute;
      inset: 0;
      pointer-events: none;
      z-index: 10;
    }

    .particle {
      position: absolute;
      width: 4px;
      height: 4px;
      border-radius: 50%;
      background: rgba(255, 255, 255, 0.6);
    }
  </style>
</head>
<body>
  <!-- Root composition -->
  <div data-composition-id="post-production"
       data-width="1920" data-height="1080" data-fps="30"
       data-duration="VIDEO_DURATION">

    <!-- Track 0: 原始视频（或去背景后的视频 + 新背景） -->
    <video id="main-video"
           data-start="0" data-duration="VIDEO_DURATION" data-track-index="0"
           src="input.mp4"
           style="position:absolute;inset:0;width:100%;height:100%;object-fit:cover;">
    </video>

    <!-- Track 1: 字幕层 -->
    <div id="cap-1" class="caption" data-start="0.5" data-duration="2.3" data-track-index="1">
      大家好，欢迎来到今天的分享
    </div>
    <div id="cap-2" class="caption" data-start="2.8" data-duration="2.0" data-track-index="1">
      今天我们聊一聊 AI 编程
    </div>
    <!-- ... 更多字幕 ... -->

    <!-- Track 2: 特效叠加层 -->
    <div id="effects" class="effect-layer"
         data-start="0" data-duration="VIDEO_DURATION" data-track-index="2">
      <!-- 粒子/光效/图标等 -->
    </div>

    <!-- Track 3: BGM -->
    <audio data-start="0" data-duration="VIDEO_DURATION" data-track-index="3"
           src="bgm.wav" data-volume="0.25"></audio>

  </div>

  <script src="https://cdn.jsdelivr.net/npm/gsap@3/dist/gsap.min.js"></script>
  <script>
    // 字幕入场动画
    const captionTl = gsap.timeline({ paused: true });
    document.querySelectorAll('.caption').forEach(cap => {
      const start = parseFloat(cap.dataset.start);
      captionTl.fromTo(cap,
        { opacity: 0, y: 20 },
        { opacity: 1, y: 0, duration: 0.3 },
        start
      );
      captionTl.to(cap,
        { opacity: 0, duration: 0.2 },
        start + parseFloat(cap.dataset.duration) - 0.2
      );
    });

    window.__timelines = window.__timelines || {};
    window.__timelines["post-production"] = captionTl;
  </script>
</body>
</html>

去背景 + 换背景模板

<!-- Track 0: 新背景 -->
<div id="new-bg" data-start="0" data-duration="VIDEO_DURATION" data-track-index="0"
     style="position:absolute;inset:0;background:linear-gradient(135deg, #1a1a2e 0%, #16213e 50%, #0f3460 100%);">
  <!-- 可放动态网格、粒子等背景动画 -->
</div>

<!-- Track 1: 去背景后的人物 -->
<video id="person" data-start="0" data-duration="VIDEO_DURATION" data-track-index="1"
       src="transparent.webm"
       style="position:absolute;bottom:0;left:50%;transform:translateX(-50%);height:90%;object-fit:contain;">
</video>

<!-- Track 2: 字幕 -->
<!-- ... -->

画中画 (PiP) 模板

<!-- Track 0: 主画面（全屏） -->
<video id="main" data-start="0" data-duration="VIDEO_DURATION" data-track-index="0"
       src="main-content.mp4"
       style="position:absolute;inset:0;width:100%;height:100%;object-fit:cover;">
</video>

<!-- Track 1: 画中画（右下角小窗） -->
<video id="pip" data-start="0" data-duration="VIDEO_DURATION" data-track-index="1"
       src="talking-head.mp4"
       style="position:absolute;bottom:40px;right:40px;width:360px;height:360px;
              border-radius:50%;object-fit:cover;border:3px solid rgba(255,255,255,0.8);
              box-shadow:0 4px 20px rgba(0,0,0,0.3);">
</video>

<!-- PiP 变体：左下角矩形 -->
<!--
<video id="pip-rect" ...
       style="position:absolute;bottom:40px;left:40px;width:480px;height:270px;
              border-radius:12px;object-fit:cover;border:2px solid rgba(255,255,255,0.5);">
</video>
-->

分屏布局模板

<!-- 左右分屏 50/50 -->
<video id="left" data-start="0" data-duration="VIDEO_DURATION" data-track-index="0"
       src="screen-recording.mp4"
       style="position:absolute;left:0;top:0;width:50%;height:100%;object-fit:cover;">
</video>
<video id="right" data-start="0" data-duration="VIDEO_DURATION" data-track-index="0"
       src="talking-head.mp4"
       style="position:absolute;right:0;top:0;width:50%;height:100%;object-fit:cover;">
</video>
<!-- 中间分割线 -->
<div data-start="0" data-duration="VIDEO_DURATION" data-track-index="1"
     style="position:absolute;left:50%;top:0;width:2px;height:100%;background:rgba(255,255,255,0.3);transform:translateX(-50%);">
</div>

Phase 4: 音频混音策略

基本混音（BGM + 原声）

# 1. 渲染视频（无音频）
npx hyperframes render --non-interactive

# 2. 提取原始口播音频
ffmpeg -i input.mp4 -vn -acodec pcm_s16le -ar 44100 -ac 2 voice.wav

# 3. 混合：原声为主，BGM 为辅
ffmpeg -i voice.wav -i bgm.wav -filter_complex \
  "[0:a]volume=1.0[voice];[1:a]volume=0.25[bgm];[voice][bgm]amix=inputs=2:duration=first" \
  -ac 2 -ar 44100 mixed_audio.wav

# 4. 合并视频 + 混音
ffmpeg -i rendered_video.mp4 -i mixed_audio.wav -c:v copy -c:a aac -shortest final.mp4

高级：BGM 自动避让（Ducking）

当人声出现时 BGM 自动降低音量，人声停顿时 BGM 恢复：

# 使用 sidechaincompress 实现 ducking
ffmpeg -i voice.wav -i bgm.wav -filter_complex \
  "[1:a]volume=0.35[bgm_vol];\
   [bgm_vol][0:a]sidechaincompress=threshold=0.02:ratio=4:attack=200:release=1000[bgm_ducked];\
   [0:a][bgm_ducked]amix=inputs=2:duration=first[out]" \
  -map "[out]" -ac 2 -ar 44100 mixed_ducked.wav

参数说明：

• threshold=0.02: 人声信号强度阈值（越低越敏感）
• ratio=4: 压缩比（4:1 表示 BGM 降到原来 1/4）
• attack=200: 压缩启动时间 200ms（避免突然降低）
• release=1000: 释放时间 1000ms（人声停后 1 秒 BGM 恢复）

音量标准化

# 测量当前音量
ffmpeg -i mixed_audio.wav -af "volumedetect" -f null /dev/null

# 标准化到 -16 LUFS（适合社交媒体）
ffmpeg -i mixed_audio.wav -af "loudnorm=I=-16:TP=-1.5:LRA=11" normalized.wav

特效叠加模板库

1. 粒子飘落效果

<div id="particles" class="effect-layer" data-start="0" data-duration="VIDEO_DURATION" data-track-index="2">
  <!-- 粒子由 JS 生成 -->
</div>

<script>
function mulberry32(seed) {
  return function() {
    seed |= 0; seed = seed + 0x6D2B79F5 | 0;
    let t = Math.imul(seed ^ seed >>> 15, 1 | seed);
    t = t + Math.imul(t ^ t >>> 7, 61 | t) ^ t;
    return ((t ^ t >>> 14) >>> 0) / 4294967296;
  }
}
const rand = mulberry32(42);

// 生成粒子
const container = document.getElementById('particles');
for (let i = 0; i < 30; i++) {
  const p = document.createElement('div');
  p.className = 'particle';
  p.style.cssText = `
    left: ${rand() * 100}%;
    top: -10px;
    width: ${3 + rand() * 4}px;
    height: ${3 + rand() * 4}px;
    opacity: ${0.3 + rand() * 0.5};
    background: hsl(${200 + rand() * 60}, 80%, 70%);
  `;
  container.appendChild(p);
}

// 粒子下落动画
const particleTl = gsap.timeline({ paused: true });
container.querySelectorAll('.particle').forEach((p, i) => {
  particleTl.to(p, {
    y: 1200,
    x: `+=${(rand() - 0.5) * 200}`,
    duration: 4 + rand() * 3,
    repeat: Math.floor(VIDEO_DURATION / 6),
    ease: "none",
    delay: rand() * 3
  }, 0);
});
// 注意：将 particleTl 加入 window.__timelines
</script>

2. 底部动态信息条（Lower Third）

<div id="lower-third" class="effect-layer" data-start="2" data-duration="8" data-track-index="2">
  <div style="position:absolute;bottom:60px;left:60px;display:flex;align-items:center;gap:16px;">
    <div style="width:4px;height:48px;background:linear-gradient(180deg,#00d4ff,#7b2ff7);border-radius:2px;"></div>
    <div>
      <div style="font-family:'Inter',sans-serif;font-size:28px;font-weight:700;color:#fff;">张三</div>
      <div style="font-family:'Inter',sans-serif;font-size:20px;color:rgba(255,255,255,0.7);margin-top:4px;">高级产品经理 · 某科技公司</div>
    </div>
  </div>
</div>

<script>
const ltTl = gsap.timeline({ paused: true });
const lt = document.querySelector('#lower-third > div');
ltTl.fromTo(lt, { x: -300, opacity: 0 }, { x: 0, opacity: 1, duration: 0.5, ease: "power2.out" }, 2);
ltTl.to(lt, { x: -300, opacity: 0, duration: 0.4, ease: "power2.in" }, 9.5);
// 加入 window.__timelines
</script>

3. 光效扫描（Light Sweep）

<div id="light-sweep" class="effect-layer" data-start="0" data-duration="VIDEO_DURATION" data-track-index="2">
  <div class="sweep-bar" style="
    position:absolute;
    top:0;left:-200px;
    width:200px;height:100%;
    background:linear-gradient(90deg, transparent, rgba(255,255,255,0.08), transparent);
    transform:skewX(-15deg);
  "></div>
</div>

<script>
const sweepTl = gsap.timeline({ paused: true });
sweepTl.to('.sweep-bar', {
  x: 2200,
  duration: 3,
  repeat: Math.floor(VIDEO_DURATION / 5),
  repeatDelay: 2,
  ease: "power1.inOut"
}, 0);
</script>

4. 关键词高亮弹出

<!-- 在特定时间点弹出关键信息 -->
<div id="keyword-pop" data-start="5" data-duration="3" data-track-index="2"
     style="position:absolute;top:50%;right:80px;transform:translateY(-50%);
            font-family:'Inter',sans-serif;font-size:56px;font-weight:900;
            color:#00d4ff;text-shadow:0 0 20px rgba(0,212,255,0.5);">
  效率提升 300%
</div>

<script>
const kwTl = gsap.timeline({ paused: true });
kwTl.fromTo('#keyword-pop',
  { scale: 0, opacity: 0, rotation: -5 },
  { scale: 1, opacity: 1, rotation: 0, duration: 0.4, ease: "back.out(1.7)" },
  5
);
kwTl.to('#keyword-pop', { opacity: 0, y: -30, duration: 0.3 }, 7.5);
</script>

字幕动效选项

AI 根据视频风格自动选择最合适的字幕动效：

动效类型	适用场景	CSS/GSAP 实现
淡入淡出	正式/商务口播	opacity: 0→1→0
底部弹出	活泼/教程	y: 20→0, opacity: 0→1
逐字打字机	科技/极客风	每字 stagger 0.05s
卡拉OK高亮	重点强调	word-level 颜色切换
缩放弹入	短视频/抖音风	scale: 0.5→1, ease: back.out

卡拉OK高亮实现

<div id="cap-karaoke" class="caption" data-start="3" data-duration="2.5" data-track-index="1">
  <span class="word" data-word-start="3.0" data-word-end="3.4">今天</span>
  <span class="word" data-word-start="3.4" data-word-end="3.7">我们</span>
  <span class="word" data-word-start="3.7" data-word-end="4.1">来聊</span>
  <span class="word" data-word-start="4.1" data-word-end="4.5">AI</span>
  <span class="word" data-word-start="4.5" data-word-end="5.0">编程</span>
</div>

<style>
.word { color: rgba(255,255,255,0.5); transition: color 0.1s; }
.word.active { color: #00d4ff; text-shadow: 0 0 10px rgba(0,212,255,0.5); }
</style>

<script>
const karaokeTl = gsap.timeline({ paused: true });
document.querySelectorAll('#cap-karaoke .word').forEach(w => {
  const start = parseFloat(w.dataset.wordStart);
  karaokeTl.to(w, { className: "+=active", duration: 0.01 }, start);
});
</script>

逐字打字机实现

const typeTl = gsap.timeline({ paused: true });
const capEl = document.getElementById('cap-type');
const chars = capEl.textContent.split('');
capEl.textContent = '';
chars.forEach(ch => {
  const span = document.createElement('span');
  span.textContent = ch;
  span.style.opacity = '0';
  capEl.appendChild(span);
});
typeTl.to(capEl.querySelectorAll('span'), {
  opacity: 1,
  stagger: 0.05,
  duration: 0.01
}, parseFloat(capEl.dataset.start));

快速口播加工模式

当用户只说 "帮我这段口播加字幕和特效" 时，AI 自动执行以下完整流程：

用户提供 MP4 → 
  ① ffprobe 检查素材参数
  ② 标准化为 1080p 30fps（如需）
  ③ hyperframes transcribe --model medium
  ④ 自动字幕分组（中文 14字/行，按标点断句）
  ⑤ 选择字幕动效（默认：底部弹出 + 半透明底板）
  ⑥ 选择特效层（默认：轻微粒子 + 底部信息条）
  ⑦ 编写 Composition HTML
  ⑧ hyperframes render
  ⑨ 音频处理（默认保留原片完整音频；仅在用户要求时混入 BGM）
  ⑩ 自检管线 Phase B-D
  ⑪ 交付 final.mp4

AI 默认选择（用户未指定时）：

• 字幕样式：底部居中，42px，白字 + 半透明黑底
• 字幕动效：淡入淡出
• 特效层：轻微光效扫描（不抢视觉焦点）
• BGM：无（除非用户要求）
• 画面处理：保持原始画面不去背景

用户可随时覆盖任何默认选择。

口播后期自检追加项

在标准 Phase A-D 自检管线之外，口播后期处理需额外检查：

#	检查项	Pass 标准
P1	字幕与语音同步	字幕出现时间 ≤ 语音开始后 0.2s
P2	字幕无遮挡关键画面	字幕区域（bottom 100px）无人脸/关键信息
P3	原视频音画同步	渲染后口型与音频匹配
P4	特效层不干扰主内容	特效透明度 ≤ 0.3，不遮挡人物/字幕
P5	BGM ducking 生效	人声段 BGM 降到 -20dB 以下
P6	视频时长完整	output duration ≥ input duration - 0.1s

---

Sub-Compositions

For complex videos, split into separate HTML files:

<!-- In index.html -->
<div id="scene-1" data-composition-id="intro"
     data-composition-src="compositions/intro.html"
     data-start="0" data-duration="5" data-track-index="1"></div>

Sub-composition files use <template> wrapper (main index.html does NOT):

<template id="intro-template">
  <div data-composition-id="intro" data-width="1920" data-height="1080">
    <!-- content, style, script -->
  </div>
</template>

Variables (Parametrized Compositions)

Declare on <html> root, read with window.__hyperframes.getVariables():

<html data-composition-variables='[
  {"id":"title","type":"string","label":"Title","default":"Hello"},
  {"id":"accent","type":"color","label":"Accent Color","default":"#ff6b35"}
]'>

Override at render: npx hyperframes render --variables '{"title":"Q4 Report"}'

Iterative Fix Decision Tree

When the user requests modifications after initial delivery, follow this decision tree to minimize re-work:

Fix Classification

Change type	Scope	Actions required
Text/data change	Single scene content	Edit HTML → Lint → Re-render → Re-mux audio → Verify
Layout/position fix	CSS class or inline style	Edit CSS → Lint → Re-render → Re-mux audio → Verify
Color/style change	CSS variables or colors	Edit CSS → Lint → Re-render → Re-mux audio → Verify
Animation timing	GSAP parameters	Edit JS → Lint → Re-render → Re-mux audio → Verify
Add/remove scene	Structure change	Edit HTML+JS → Recalculate all timings → Lint → Re-render → Re-mux → Verify
BGM style change	Audio only	Only when user requested BGM change: regenerate/replace BGM → Re-mux only (skip re-render) → Verify audio
Duration change	Everything	Full rebuild required

Quick-Fix Workflow (for text/layout/color/animation changes)

1. Identify affected scene(s) — read current HTML
2. Make targeted edit(s) — ONLY touch affected parts
3. npx hyperframes lint — must pass
4. npx hyperframes render --quality draft — fast preview check
5. If draft looks good → render --quality standard
6. Re-mux the approved audio source (original_audio.wav for existing videos, bgm.wav only for generated-BGM projects)
7. Verify final MP4 (Phase C+D checks)
8. Deliver

Key optimization: DO NOT regenerate or replace audio unless the user requests a music style change or duration change. For existing source videos, reuse the extracted original_audio.wav; for generated-BGM projects, reuse the existing approved bgm.wav across layout/content fixes.

When to Use --quality draft First

ALWAYS render draft quality first when:

• Fixing layout issues (verify position before full render)
• User reported visual problems (confirm fix before spending 10+ minutes)
• Making multiple iterative adjustments (draft → confirm → standard)

Draft renders at ~3x speed. Only proceed to standard after visual confirmation.

Batch Fix Strategy

When user reports multiple issues at once:

1. Collect ALL reported issues
2. Fix ALL issues in a single pass (edit HTML once)
3. Lint once
4. Render once (not once per fix)
5. Verify all fixes in the rendered output

NEVER render between each individual fix — that wastes 10+ minutes per cycle.

---

Chinese Content Adaptation (Soft Default — 中文适配参考)

以下规则是中文视频的经验参考，帮助 AI 在用户未给出详细排版指示时做出合理的中文布局决策。当用户的提示词有明确的排版/密度/风格要求时，以用户要求为准。

Text Density Reference

Scene duration	建议中文字数 (同时在屏)	建议内容项数
3 seconds	60-80 字	3-4 items
4 seconds	80-120 字	4-6 items
5+ seconds	120-160 字	6-8 items

自检关注点: 如果某场景文字过密导致不可读（字号 ≤ 28px 且停留 ≤ 3s），AI 应自动拆分——这是可读性问题，不是风格限制。

Chinese Typography CSS (推荐实践)

/* Line breaking — keep Chinese words together */
.content-area, .content-col, .glass-card {
  word-break: keep-all;        /* prevent mid-word breaks */
  overflow-wrap: break-word;   /* break only at natural points */
  line-break: strict;          /* no punctuation at line start */
}

/* Chinese line height — wider than English */
p, span, .card-desc { line-height: 1.6; }   /* body text */
h2, h3 { line-height: 1.3; }                /* titles */

/* Mixed CJK + Latin spacing */
.mixed-text { text-spacing-trim: space-all; } /* if supported */
/* Fallback: manually add thin space between Chinese and numbers/English */

Chinese-English Mixed Content Conventions

Pattern	Example	Rule
Number + Chinese unit	128亿美元	No space between number and Chinese
English brand + Chinese	GitHub Copilot 工具	Space between English and Chinese
Percentage	156% or 156％	Use half-width % (more compact)
Punctuation	，、。；	Use full-width Chinese punctuation in body text
Data labels	市场规模：	Use full-width colon ： in Chinese context
Card titles	核心技术突破	No trailing punctuation on card titles

Chinese Layout Adjustments (参考)

Chinese characters are wider than Latin characters. 当 AI 自行规划布局时可参考：

Element	English width	Chinese adjustment
Card title	40-44px	38-42px (reduce 2px)
Body text	34-38px	32-36px (reduce 2px)
Card width	360px	380-400px (increase 20-40px)
Line chars	~40 chars/line	~18-22 中文字/行

Scene Content Text Templates (灵感参考，非必须)

数据展示场景:
  主数据: "128亿" (数字 88-96px + 单位 42px)
  标签: "市场规模" (38px, opacity 0.7)

卡片场景:
  图标: 64-72px emoji/SVG
  卡片标题: "核心技术突破" (40px, bold)
  描述文字: "一句话说明功能或数据" (32-34px, opacity 0.8)
  底部标注: "具体数据或来源" (28px, opacity 0.6)

列表场景:
  序号: "01" (56px, accent color)
  内容: "一行描述，不超过25字" (36px)

NOTE: 以上仅为参考模板。AI 完全可以使用不同的信息层级、卡片结构或数据展示方式。

---

Vertical Video Safe Zones (P1 — 1080×1920)

Layout System for 9:16 Vertical Videos

┌───────────────────────┐
│ Top Safe: 120px       │ ← Platform UI (status bar)
├───────────────────────┤
│ Title Zone:           │
│ top 160px, h ≤ 200px  │
├───────────────────────┤
│                       │
│ Content Zone:         │
│ top 400px ~ bot 1600px│ ← Available: 1200px
│                       │
├───────────────────────┤
│ Bottom Safe: 320px    │ ← Platform UI (controls, comments)
└───────────────────────┘

Vertical Video Typography Scale

Element	Safe range	Hard max
Cover mega title	120-160px	180px
Scene title	80-100px	120px
Subtitle	48-64px	72px
Body text	36-48px	56px
Card title	40-52px	56px
Data number	96-128px	140px
Icon/emoji	72-108px	128px

Vertical Content Rules

• Max 2 cards horizontally (full width), prefer single-column stacking
• Card width: 90% container (≈ 972px)
• Scene transitions: faster pacing (2-3s per scene typical for Reels/TikTok)
• Bottom 320px always clear (platform overlays on mobile)
• Horizontal safe margin: ≥ 60px (narrower than landscape)

---

Icon & Emoji Strategy (P1)

Recommended Approach Priority

1. Unicode Emoji (first choice for most cases)

   • ✅ Renders consistently in Chrome Headless
   • ✅ No external dependencies
   • ✅ Supports all common categories
   • ⚠️ Style varies slightly across platforms (but video rendering uses Chrome's Noto Emoji)

2. Inline SVG (when custom icons needed)

   • ✅ Pixel-perfect control
   • ✅ Animatable with GSAP
   • ✅ Color matches theme exactly
   • ⚠️ Increases HTML file size

3. CSS-drawn shapes (for simple geometric icons)

   • ✅ No external resources
   • ✅ Fully animatable
   • ⚠️ Limited to simple shapes

PROHIBITED approaches

• ❌ Font Awesome / Material Icons CDN (render environment may not load)
• ❌ External image URLs (network dependency = unreliable)
• ❌ Icon font @import (same issue as Google Fonts)

Common Tech Video Icon Set (copy-paste ready)

Categories:
💻 编程/开发   🚀 发布/增长   📊 数据/图表   🔧 工具/设置
🎯 目标/聚焦   ⚡ 性能/速度   🔒 安全/隐私   🌐 网络/全球
📱 移动端      🤖 AI/机器人   🎮 游戏        🏗️ 架构/构建
💡 创新/灵感   📈 增长/趋势   🛡️ 防护/安全   ⏱️ 时间/效率

Specific use cases:
Web开发: 🌐    移动应用: 📱    AI/ML: 🤖    游戏: 🎮
网络安全: 🔒   数据科学: 📊    代码: 💻     部署: 🚀
效率: ⚡       质量: ✅        风险: ⚠️     趋势: 📈

Icon Sizing Rules

/* Standard icon in card */
.card-icon { font-size: 64px; line-height: 1; }

/* Small inline icon */
.inline-icon { font-size: 48px; vertical-align: middle; }

/* Feature highlight icon */
.feature-icon { font-size: 72px; }

/* NEVER exceed these for icons: */
/* Horizontal video: 96px max */
/* Vertical video: 128px max */

Custom SVG Icon Template

<!-- Reusable SVG icon pattern for tech videos -->
<svg width="64" height="64" viewBox="0 0 64 64" fill="none">
  <circle cx="32" cy="32" r="28" stroke="currentColor" stroke-width="2" opacity="0.3"/>
  <path d="M20 32 L28 40 L44 24" stroke="currentColor" stroke-width="3" stroke-linecap="round"/>
</svg>

---

Render Time Estimation (P1)

Estimated Render Duration by Quality

Quality	Speed ratio	10s video	30s video	42s video	60s video
draft	~6fps	~2 min	~5 min	~7 min	~10 min
standard	~3fps	~3 min	~10 min	~14 min	~20 min
high	~1.5fps	~7 min	~20 min	~28 min	~40 min

Times are approximate. Complex scenes (many particles, gradients, blur filters) render slower.

Factors That Increase Render Time

Factor	Impact	Mitigation
backdrop-filter: blur()	+30-50%	Limit to 3-4 elements max
> 50 particles	+20%	Cap at 40, reduce size
Multiple box-shadows	+15%	Use single subtle shadow
60fps (vs 30fps)	+100%	Use 30fps unless requested
Large video backgrounds	+40%	Use gradient/CSS backgrounds instead

User Communication Template

Before starting a render, inform the user:

"开始渲染 [质量] 品质视频（[时长]秒），预计需要 [X-Y] 分钟。
渲染期间我会持续检查进度，完成后立即进行音频合成和质量验证。"

Render Strategy Decision

用户要求"快速看一下效果" → --quality draft
用户要求"正式版/最终版"  → --quality standard
用户明确说"最高画质"     → --quality high --fps 60
修复布局问题验证        → --quality draft (先确认再正式渲染)

---

Performance Optimization (P2)

DOM Element Limits

Complexity level	Max DOM elements	Max particles	Max cards	Render impact
Light	< 200	20	3-4	Normal speed
Medium	200-500	40	6-8	+20% time
Heavy	500-1000	60	10-12	+50% time
Danger zone	> 1000	> 80	> 15	May crash Chrome

Optimization Techniques

/* Use will-change for animated elements (Chrome optimization) */
.scene-wrapper, .glass-card, .particle {
  will-change: transform, opacity;
}

/* Reduce paint complexity */
.particle {
  border-radius: 50%;
  /* Use background-color, NOT box-shadow for particles */
  background: currentColor;
}

/* GPU-accelerated properties only */
/* PREFER: transform, opacity */
/* AVOID: width, height, top, left, margin, padding, box-shadow (triggers layout) */

When Chrome Crashes During Render

1. Reduce --workers to 1: npx hyperframes render --workers 1
2. Remove excessive particles (cap at 30)
3. Replace backdrop-filter: blur() with pre-blurred gradient backgrounds
4. Remove multiple box-shadow layers
5. If still crashing: split into shorter sub-compositions, render separately, concatenate with FFmpeg

Render Timeout Handling

If render exceeds 20 minutes for a ≤ 60s video:

1. Check if Chrome process is still alive (ps aux | grep chrome)
2. If frozen: kill and retry with --workers 1 --quality draft
3. If draft succeeds: the standard render had a resource issue → simplify complex scenes
4. Report specific scene if identifiable (check last rendered frame number in output)

---

Multi-Resolution Adaptation (P2)

Supported Canvas Sizes

Aspect ratio	Resolution	Use case	Init config
16:9 横版	1920×1080	YouTube, 公众号, B站	data-width="1920" data-height="1080"
9:16 竖版	1080×1920	抖音, Reels, 视频号	data-width="1080" data-height="1920"
1:1 正方形	1080×1080	Instagram Feed, 小红书	data-width="1080" data-height="1080"
4:5 竖版	1080×1350	Instagram Feed (推荐)	data-width="1080" data-height="1350"
4:3 标准	1440×1080	演示文稿风格	data-width="1440" data-height="1080"

Per-Resolution Safe Zone Quick Reference

Resolution	Title top	Content top	Content bottom	Side margin
1920×1080	50px	240px	980px	120px
1080×1920	160px	400px	1600px	60px
1080×1080	50px	200px	980px	80px
1080×1350	80px	260px	1200px	80px

Resolution-Specific Typography Scale

Element	1920×1080	1080×1920	1080×1080
Mega title	96-120px	120-160px	80-100px
Scene title	64-84px	80-100px	56-72px
Body text	28-42px	36-48px	28-38px
Card title	32-44px	40-52px	30-40px
Icon	56-96px	72-108px	48-80px

Cross-Resolution Card Layouts

1920×1080 (横版):
  3 horizontal cards: 480px each + 36px gap
  2×3 grid: 520px × 300px cards
  
1080×1920 (竖版):
  2 horizontal cards: 480px each + 24px gap
  1-column stack: 920px wide cards
  
1080×1080 (正方形):
  2 horizontal cards: 460px each + 24px gap
  2×2 grid: 460px × 340px cards

---

Delivery & File Management (P2)

File Naming Convention

{project_name}_{version}_{quality}.mp4

Examples:
  ai_coding_video_v1_standard.mp4      ← first delivery
  ai_coding_video_v2_standard.mp4      ← after layout fix
  ai_coding_video_v3_standard.mp4      ← after content update
  ai_coding_video_final_standard.mp4   ← user-approved final

Intermediate File Cleanup

After user approves a final version:

# Keep: final approved MP4, source HTML, bgm.wav
# Remove: intermediate renders
rm -f render_v1.mp4 render_v2.mp4 render_v3.mp4
rm -f final_ai_coding_v1.mp4 final_ai_coding_v2.mp4
# Keep: final_ai_coding_v3.mp4 (or rename to _final)

Rule: Never auto-delete intermediate files without user confirmation. Ask: "是否清理中间渲染文件？只保留最终版本。"

Delivery Checklist

Before calling deliver_attachments:

1. ✅ File is the post-mux version (not raw render)
2. ✅ Audio verified (Phase C checks passed)
3. ✅ File size reasonable (1080p 42s ≈ 5-15MB for standard quality)
4. ✅ Filename clearly indicates version

What to Preserve for Future Edits

Always keep in project directory:

• index.html — source composition (for re-editing)
• original_audio.wav — extracted original audio for existing-video edits, when source audio must be preserved
• bgm.wav — approved BGM audio only when the project uses generated or external BGM
• Customized copies of bundled helper scripts only if they were modified for this project (scripts/verify_audio.py, scripts/gen_bgm.py remain available in the Skill package)
• Final approved .mp4 — the delivered product

Quality Checklist (Final Gate — 交付前必过)

Before delivering any video, execute the full AI Self-Check Pipeline (Section 9). The checklist adapts to prompt mode:

硬约束检查（两种模式都必须通过）

1. Lint pass: npx hyperframes lint → 0 errors. (H-all)
2. No prohibited code: No Math.random, no banned fonts, no inline top:% overrides. (H1, H2, H7)
3. Deterministic rendering: No non-deterministic APIs, GSAP repeat uses Math.floor. (H1, H8)
4. Root duration alignment: data-duration on root = last scene end time.
5. Audio full coverage (≥30s video): FFmpeg post-mux → verify duration ≥ video, no silence in last 12s, RMS -15~-20dB. (H9)
6. Visual integrity: All content within canvas bounds, no clipping, no unreadable overlaps. (H6)
7. Final delivery: Only deliver the verified post-mux MP4. Never send intermediate renders.

视觉质量检查（两种模式都必须通过）

8. Frame spot-check: Extract mid-scene screenshots → verify no overlap/overflow/clipping.
9. Readability: 关键文字可读（字号 + 对比度 + 停留时间足够）.
10. Transitions: 场景之间有过渡，无未经用户要求的跳切.

提示词合规检查（详细提示词模式）

11. Prompt compliance: Content points, visual style, color palette (user-specified, never overridden), BGM style, duration all match original prompt.

布局参考检查（仅使用软默认布局时）

12. Standard layout zone: Content ≥ 240px, ≤ 980px; Title zone not overlapped.
13. Card density: Compute content_height vs available_height. If overflow → reduce/split.
14. Typography scale: Text/icon sizes within reference range hard max.

Note: Items 12-14 仅在 AI 使用了标准布局骨架时检查。如果 AI 根据用户提示词或创意判断使用了自定义布局，这些项被替换为 Item 8 (frame spot-check) 的通过即可。

已知限制

#	限制	影响	应对方式
1	HyperFrames 内置音频 ≥32s 截断	长视频音频不完整	FFmpeg 后置合成完整音频 (H9)
2	仅支持 Inter/JetBrains Mono/Roboto 字体	中文字体无法自定义	依赖 sans-serif 系统回退
3	不支持 @import url() 引入字体	自定义 Google Fonts 不可用	仅用 Compiler 自动解析的字体
4	渲染需 Headless Chrome + FFmpeg	无法在纯容器环境直接运行	需安装完整依赖或使用 Docker
5	非确定性 API 会导致帧不一致	Math.random() 等使渲染结果不可复现	使用 mulberry32 seeded PRNG
6	单次渲染 DOM 复杂度上限 ~500 节点	超出可能导致 Chrome crash	拆分场景或简化 DOM
7	竖版视频 (1080×1920) 为实验性支持	部分动画比例需手动调整	使用竖版安全区参考表
8	渲染速度受机器性能影响	42s 视频约需 3-8 分钟渲染	开发阶段用 draft 模式预览

---

Troubleshooting

npx hyperframes doctor         # Check environment
npx hyperframes browser        # Manage bundled Chrome
npx hyperframes info           # Version details

Common issues:

• "FFmpeg not found" → brew install ffmpeg
• "Chrome not found" → npx hyperframes browser to download
• Render hangs → Check for repeat: -1 in timelines
• Empty frames → Ensure window.__timelines is registered synchronously

References

• references/composition-rules.md — Full data-attribute spec, track rules, sub-composition format
• references/animation-guide.md — GSAP patterns, easing reference, transition catalog
• references/caption-patterns.md — Caption styling, word-level sync, karaoke effects