REQUIRED: Load Remotion Best Practices First

This skill depends on remotion-best-practices (official Remotion best practices). You MUST invoke it before proceeding:

Skill tool: skill="remotion-best-practices"

Video Podcast Maker

Quick Start

打开 Claude Code，直接说："帮我制作一个关于 [你的主题] 的 B站视频播客"

Auto Update Check

Claude behavior: 每次 skill 被调用时，自动检查是否有新版本：

timeout 5 git -C ~/.claude/skills/video-podcast-maker fetch --quiet 2>/dev/null || true
LOCAL=$(git -C ~/.claude/skills/video-podcast-maker rev-parse HEAD 2>/dev/null)
REMOTE=$(git -C ~/.claude/skills/video-podcast-maker rev-parse origin/main 2>/dev/null)
if [ -n "$LOCAL" ] && [ -n "$REMOTE" ] && [ "$LOCAL" != "$REMOTE" ]; then
  echo "UPDATE_AVAILABLE"
else
  echo "UP_TO_DATE"
fi

• 有更新时：使用 AskUserQuestion 提示用户 "video-podcast-maker skill 有新版本可用，是否更新？"
  • 是（推荐） → 执行 git -C ~/.claude/skills/video-podcast-maker pull
  • 否 → 继续使用当前版本
• 已是最新：静默继续，不打扰用户

Prerequisites (One-time Setup)

0.1 环境检查清单

0.2 API 密钥

# Azure Speech (必需) - 添加到 ~/.zshrc
export AZURE_SPEECH_KEY="your-azure-speech-key"
export AZURE_SPEECH_REGION="eastasia"

# 验证
echo $AZURE_SPEECH_KEY  # 应显示你的密钥

获取方式：Azure 门户 → 创建"语音服务"资源

0.3 Python 依赖

pip install azure-cognitiveservices-speech requests

0.4 Remotion 项目设置

# 创建 Remotion 项目（如已有则跳过）
npx create-video@latest my-video-project
cd my-video-project
npm i  # 安装依赖

# 验证
npx remotion studio  # 应打开浏览器预览

0.5 快速验证

# 一键检查所有依赖
echo "=== 环境检查 ===" && \
node -v && \
python3 --version && \
ffmpeg -version 2>&1 | head -1 && \
[ -n "$AZURE_SPEECH_KEY" ] && echo "✓ AZURE_SPEECH_KEY 已设置" || echo "✗ AZURE_SPEECH_KEY 未设置"

Overview

Automated pipeline to create professional Bilibili (B站) 横屏知识视频 from a topic.

目标平台：B站横屏视频 (16:9)

• 分辨率：3840×2160 (4K) 或 1920×1080 (1080p)
• 风格：简约纯白（默认）

技术栈： Claude + Azure TTS + Remotion + FFmpeg

适用场景

输出规格

Execution Modes

Claude behavior: At the start of the workflow, detect the user's intent:

• If user says "帮我制作..." / "make a video about..." with no special instructions → Auto Mode
• If user says "我想自己控制每个步骤" / mentions interactive/manual → Interactive Mode
• Default: Auto Mode

Auto Mode (Default)

Runs the full pipeline with sensible defaults. Only 1 mandatory stop:

1. Step 10: After preview render (720p) — user reviews preview.mp4, confirms before 4K render

All other decisions use these defaults:

Users can override any default by mentioning it in their initial request:

• "帮我做一个关于AI的视频，要烧字幕" → auto mode + subtitles on
• "用深色主题，AI生成封面" → auto mode + dark theme + imagen thumbnails
• "需要截图素材" → auto mode + media collection enabled

Interactive Mode

Prompts the user at each decision point. Activated by:

• "互动模式" / "interactive mode"
• "我想自己选择每个选项"
• User explicitly requests control over specific steps

Workflow State & Resume

Claude behavior: Automatically persist workflow progress for error recovery.

State File

Each video project maintains videos/{name}/workflow_state.json:

{
  "video_name": "ai-agents-explained",
  "mode": "auto",
  "started_at": "2026-03-16T10:30:00",
  "current_step": 8,
  "steps": {
    "1": { "status": "completed", "completed_at": "2026-03-16T10:31:00" },
    "2": { "status": "completed", "completed_at": "2026-03-16T10:35:00" },
    "8": { "status": "failed", "error": "AZURE_SPEECH_KEY not set" }
  }
}

Auto-Resume

Claude behavior: When the skill is invoked:

1. Check if videos/*/workflow_state.json exists for any in-progress video
2. If found, report status and ask: "检测到未完成的视频项目 {name}，当前在第 {N} 步。是否继续？"
   • 继续 → Resume from the failed/incomplete step
   • 重新开始 → Reset state, start from Step 1
   • 新视频 → Start a different video, keep old state
3. If not found, start fresh

Step Lifecycle

Each step follows this pattern:

1. Update state: status: "in_progress"
2. Execute step
3. On success: status: "completed", record completed_at
4. On failure: status: "failed", record error message
5. On skip (auto mode): status: "skipped"

Manual Resume

Users can explicitly resume:

• "继续上次的视频" → find latest workflow_state.json, resume
• "从第8步开始" → resume from Step 8 (validate prior steps' outputs exist)

Technical Rules

以下是视频制作的技术硬约束，其他视觉设计和布局由 Claude 根据内容自由发挥：

Visual Design

Full design guide: Read DESIGN_GUIDE.md when working on Step 9 (Remotion composition). Claude MUST load DESIGN_GUIDE.md before creating or modifying any Remotion components.

文件路径与命名规范

目录结构

project-root/                           # Remotion 项目根目录
├── src/remotion/                       # Remotion 源码
│   ├── compositions/                   # 视频 Composition 定义
│   ├── Root.tsx                        # Remotion 入口
│   └── index.ts                        # 导出
│
├── public/media/{video-name}/          # 素材目录 (Remotion staticFile() 可访问)
│   ├── {section}_{index}.{ext}         # 通用素材
│   ├── {section}_screenshot.png        # 网页截图
│   ├── {section}_logo.png              # Logo
│   ├── {section}_web_{index}.{ext}     # 网络图片
│   └── {section}_ai.png               # AI 生成图片
│
├── videos/{video-name}/                # 视频项目资产 (非 Remotion 代码)
│   ├── workflow_state.json             # Workflow progress (auto-managed)
│   ├── topic_definition.md             # Step 1: 主题定义
│   ├── topic_research.md               # Step 2: 研究资料
│   ├── podcast.txt                     # Step 4: 旁白脚本
│   ├── media_manifest.json             # Step 5: 素材清单
│   ├── publish_info.md                 # Step 6+13: 发布信息
│   ├── podcast_audio.wav               # Step 8: TTS 音频
│   ├── podcast_audio.srt               # Step 8: 字幕文件
│   ├── timing.json                     # Step 8: 时间轴
│   ├── thumbnail_*.png                 # Step 7: 封面
│   ├── output.mp4                      # Step 10: Remotion 输出
│   ├── video_with_bgm.mp4             # Step 11: 添加 BGM
│   ├── final_video.mp4                 # Step 12: 最终输出
│   └── bgm.mp3                         # 背景音乐
│
└── remotion.config.ts                  # Remotion 配置

⚠️ 重要: Remotion 渲染时必须指定完整输出路径，否则默认输出到 out/:

npx remotion render src/remotion/index.ts CompositionId videos/{name}/output.mp4

命名规则

视频名称 {video-name}: 全小写英文，连字符分隔（如 reference-manager-comparison）

章节名称 {section}: 全小写英文，下划线分隔，与 [SECTION:xxx] 一致

缩略图命名 (⚠️ 16:9 和 4:3 都是必须的，B站不同位置使用不同比例):

渲染前后文件操作

# 渲染前
cp videos/{name}/podcast_audio.wav videos/{name}/timing.json public/
[ -f videos/{name}/media_manifest.json ] && cp videos/{name}/media_manifest.json public/

# 渲染后清理
rm -f public/podcast_audio.wav public/timing.json public/media_manifest.json
rm -rf public/media/{name}

Workflow

Progress Tracking

在 Step 1 开始时：

1. Create videos/{name}/workflow_state.json with initial state
2. 使用 TaskCreate 按以下列表逐条创建 tasks，每步开始时 TaskUpdate 为 in_progress，完成后标记 completed
3. Each step completion updates BOTH workflow_state.json AND TaskUpdate

 1. Define topic direction (brainstorming) → topic_definition.md
 2. Research topic → topic_research.md
 3. Design video sections (5-7 chapters)
 4. Write narration script → podcast.txt
 5. Collect media assets → media_manifest.json
 6. Generate publish info (Part 1) → publish_info.md
 7. Generate thumbnails (16:9 + 4:3) → thumbnail_*.png
 8. Generate TTS audio → podcast_audio.wav, timing.json
 9. Create Remotion composition + Studio preview
10. Render 4K video → output.mp4
11. Mix background music → video_with_bgm.mp4
12. Add subtitles (optional) → final_video.mp4
13. Complete publish info (Part 2) → chapter timestamps
14. Verify output & cleanup

Validation Checkpoints

After Step 8 (TTS):

• podcast_audio.wav exists and plays correctly
• timing.json has all sections with correct timestamps
• podcast_audio.srt encoding is UTF-8

After Step 10 (Render):

• output.mp4 resolution is 3840x2160
• Audio-video sync verified
• No black frames

Step 0: Load User Preferences

Claude behavior: 自动执行，无需用户干预

1. 检查 user_prefs.json 是否存在于 skill 目录
2. 如果不存在，从 user_prefs.template.json 复制创建
3. 读取用户偏好并在后续步骤中应用

SKILL_DIR="~/.claude/skills/video-podcast-maker"
PREFS_FILE="$SKILL_DIR/user_prefs.json"
TEMPLATE_FILE="$SKILL_DIR/user_prefs.template.json"

if [ ! -f "$PREFS_FILE" ]; then
  cp "$TEMPLATE_FILE" "$PREFS_FILE"
  echo "✓ 首次使用，已创建默认偏好配置"
fi

4. 在 Step 1 开始时，告知用户当前应用的偏好（如有自定义）：

"根据您的偏好设置：
 - 主题: [theme]
 - 字号缩放: [scalePreference]x
 - 语速: [tts.rate]

如需调整请随时告诉我，或说「显示偏好设置」查看详情。"

Step 1: Define Topic Direction

Auto mode: Infer all decisions from the user's topic description. Use sensible defaults (目标受众: general, 视频定位: 科普入门, 视频风格: 轻松专业, 时长: 中 3-7分钟). Save directly to videos/{name}/topic_definition.md without prompting.

Interactive mode: 使用 brainstorming skill 逐项确认：

1. 目标受众: 技术开发者 / 普通用户 / 学生 / 专业人士
2. 视频定位: 科普入门 / 深度解析 / 新闻速报 / 教程实操
3. 内容范围: 历史背景 / 技术原理 / 使用方法 / 对比评测
4. 视频风格: 严肃专业 / 轻松幽默 / 快节奏
5. 时长预期: 短 (1-3分钟) / 中 (3-7分钟) / 长 (7-15分钟)

保存为 videos/{name}/topic_definition.md

Step 2: Research Topic

Use WebSearch and WebFetch. Save to videos/{name}/topic_research.md.

Step 3: Design Video Sections

Design 5-7 sections:

• Hero/Intro (15-25s)
• Core concepts (30-45s each)
• Demo/Examples (30-60s)
• Comparison/Analysis (30-45s)
• Summary (20-30s)

Content Density Selection

Before designing, assign each section a density tier based on content volume:

Topic Type Detection

基于主题关键词自动检测类别并匹配偏好：

Claude behavior: 检测到主题类别后，合并 topic_patterns[category] 到当前偏好。

Title Position

Auto mode: Use top-center (default for most video content). Interactive mode: Ask user to choose from: 顶部居中 (推荐) / 顶部左侧 / 全屏居中.

规则： 单个视频内保持标题位置一致。

Step 4: Write Narration Script

偏好应用: 根据 user_prefs.content 调整脚本风格：

• tone: professional → 使用正式用语，避免口语化
• tone: casual → 轻松口语，可加入语气词
• verbosity: concise → 每段 50-80 字
• verbosity: detailed → 每段 100-150 字

Create videos/{name}/podcast.txt with section markers:

[SECTION:hero]
大家好，欢迎来到本期视频。今天我们聊一个...

[SECTION:features]
它有以下功能...

[SECTION:demo]
让我演示一下...

[SECTION:summary]
总结一下，xxx是目前最xxx的xxx。

[SECTION:references]
本期视频参考了官方文档和技术博客。

[SECTION:outro]
感谢观看！点赞投币收藏，关注我，下期再见！

数字必须使用中文读音 - 所有数字必须写成中文，TTS 才能正确朗读：

章节说明:

• summary: 纯内容总结，不包含互动引导
• references (可选): 一句话概括参考来源
• outro: 感谢 + 一键三连引导
• 空内容的 [SECTION:xxx] 为静音章节

Duration Estimation (Dry Run)

Claude behavior: After writing podcast.txt, automatically run dry-run to estimate video duration:

python3 generate_tts.py --input videos/{name}/podcast.txt --output-dir videos/{name} --dry-run

Report estimated duration to user. If too long (>12min) or too short (<3min), suggest adjustments before proceeding to TTS.

Step 5: Collect Media Assets

Auto mode: Skip media collection (text-only animated sections). Proceed to Step 6. Interactive mode: Ask per-section media source (跳过 / 本地文件 / 网页截图 / 网络检索 / AI 生成).

If user mentioned AI images, screenshots, or specific assets in their initial request, collect those regardless of mode.

素材保存到 public/media/{video-name}/，生成 media_manifest.json。

可用素材来源：

• Unsplash / Pexels / Pixabay — 免费图片
• unDraw — 开源 SVG 插图
• Simple Icons — 品牌 SVG 图标
• Playwright — 网页截图
• imagen skill — AI 生成图片

Step 6: Generate Publish Info (Part 1)

基于 podcast.txt 生成 publish_info.md:

• 标题（数字 + 主题 + 吸引词）
• 标签（10个，含产品名/领域词/热门标签）
• 简介（100-200字）

Step 7: Generate Video Thumbnail

Auto mode: Generate Remotion thumbnails (16:9 + 4:3). Interactive mode: Ask user to choose: Remotion生成 / AI文生图(imagen skill) / 两者都生成.

⚠️ 必须生成两个比例: 16:9 (播放页) 和 4:3 (推荐流/动态)，缺一不可。9:16 仅在生成竖屏视频时需要。

npx remotion still src/remotion/index.ts Thumbnail16x9 videos/{name}/thumbnail_remotion_16x9.png
npx remotion still src/remotion/index.ts Thumbnail4x3 videos/{name}/thumbnail_remotion_4x3.png
# Optional: vertical thumbnail (only if rendering vertical video)
npx remotion still src/remotion/index.ts Thumbnail9x16 videos/{name}/thumbnail_remotion_9x16.png

Step 8: Generate TTS Audio

偏好应用: 从 user_prefs.tts 读取 backend/rate/voice 设置。

cp ~/.claude/skills/video-podcast-maker/generate_tts.py .

# Primary command (backend from user_prefs or env)
python3 generate_tts.py --input videos/{name}/podcast.txt --output-dir videos/{name}

# Resume from breakpoint (skip already synthesized parts)
python3 generate_tts.py --input videos/{name}/podcast.txt --output-dir videos/{name} --resume

# Dry run (estimate duration, no API call)
python3 generate_tts.py --input videos/{name}/podcast.txt --output-dir videos/{name} --dry-run

Backend selection via env: TTS_BACKEND=azure|cosyvoice|edge, rate via TTS_RATE="+5%".

多音字/发音校正 (SSML Phoneme)

TTS 脚本支持三种方式校正发音，优先级从高到低：

1. 内联标注 (最高优先级) - 在 podcast.txt 中直接标注：

每个执行器[zhí xíng qì]都有自己的上下文窗口
如果不合格，就打回重做[chóng zuò]

2. 项目词典 - 在 videos/{name}/phonemes.json 中定义：

{
  "执行器": "zhí xíng qì",
  "重做": "chóng zuò",
  "一行命令": "yì háng mìng lìng"
}

3. 内置词典 - 预置常见多音字（自动应用）

Outputs: podcast_audio.wav, podcast_audio.srt, timing.json

timing.json label field: Each section gets a human-readable label extracted from the first line of its content (before first punctuation, max 10 chars). This is displayed in the ChapterProgressBar component. Example: [SECTION:hero] with content "大家好，欢迎来到本期视频" → label: "大家好". Silent sections use the section name as label.

Step 9: Create Remotion Composition + Studio Preview

Claude MUST read DESIGN_GUIDE.md before this step.

偏好应用: 从 user_prefs.visual 覆盖 defaultVideoProps：

• typography.* × scalePreference → 应用字号缩放
• theme: dark → 交换 backgroundColor/textColor
• primaryColor, accentColor → 直接覆盖

复制文件到 public/:

cp videos/{name}/podcast_audio.wav videos/{name}/timing.json public/

标准视频模板（必须遵循）

使用 templates/Video.tsx 作为起点，已包含完整实现（4K 缩放、章节进度条、音频集成）。

Shared infrastructure — copy only if not already present (safe for existing projects):

# Root.tsx and components are shared across all videos — don't overwrite if customized
[ ! -f src/remotion/Root.tsx ] && cp ~/.claude/skills/video-podcast-maker/templates/Root.tsx src/remotion/
[ ! -d src/remotion/components ] && cp -r ~/.claude/skills/video-podcast-maker/templates/components src/remotion/components

Per-video composition — NEVER overwrite Video.tsx. Create a unique file per video:

# Generate per-video composition: {PascalCaseName}Video.tsx
# Example: "ai-agents" → AiAgentsVideo.tsx
cp ~/.claude/skills/video-podcast-maker/templates/Video.tsx src/remotion/{PascalCaseName}Video.tsx

Then register it in Root.tsx alongside existing compositions. Each video gets its own composition file with customized section layouts, colors, and components.

Naming convention:

Components are modular — import only what you need:

import { ComparisonCard, CodeBlock, FeatureGrid, MediaSection } from "./components";

章节转场效果

模板使用 @remotion/transitions 的 TransitionSeries 实现章节间平滑过渡。

安装依赖（项目中执行）：

npm install @remotion/transitions

关键架构说明

一键三连片尾

Auto mode: Use pre-made MP4 animation (white for light theme, black for dark theme). Interactive mode: Ask user to choose: 预制 MP4 动画 (推荐) / Remotion 代码生成.

# Copy to project public directory
cp ~/.claude/skills/video-podcast-maker/assets/bilibili-triple-white.mp4 public/media/{video-name}/
# Or black background version
cp ~/.claude/skills/video-podcast-maker/assets/bilibili-triple-black.mp4 public/media/{video-name}/

// In outro section, embed with <OffthreadVideo>
import { OffthreadVideo, staticFile } from "remotion";
<OffthreadVideo src={staticFile("media/{video-name}/bilibili-triple-white.mp4")} />

Preview & Quality Gate

Auto mode: Skip Studio. Proceed directly to Step 10 where a preview render (720p) is generated and Claude self-validates (duration, no black frames, audio sync). If validation passes, proceed to 4K automatically.

Interactive mode: Launch Studio for iterative visual refinement:

npx remotion studio src/remotion/index.ts

1. Launch remotion studio (real-time preview, hot reload)
2. Ask user: "预览效果满意吗？如果需要调整，请描述修改意见（例如：标题太小、背景换深色、动画太快）"
   • 满意，继续渲染 → proceed to Step 10
   • 需要修改 → apply changes, Studio hot reloads, repeat
3. Pronunciation fixes require re-running TTS (Step 8) and copying updated files to public/.

Step 9.5: Visual QA (Automated)

Claude behavior: Automatically run after Step 9 composition is created, before Step 10 render. No user prompt needed — Claude self-inspects.

Render Section Stills

Read timing.json to get each section's midpoint frame, then render a still for each:

# For each section in timing.json, render a still at its midpoint
# midpoint_frame = start_frame + (duration_frames / 2)
npx remotion still src/remotion/index.ts CompositionId videos/{name}/qa_{section_name}.png --frame {midpoint_frame}

Example with 5 sections:

npx remotion still src/remotion/index.ts CompositionId videos/{name}/qa_hero.png --frame 184
npx remotion still src/remotion/index.ts CompositionId videos/{name}/qa_features.png --frame 1566
npx remotion still src/remotion/index.ts CompositionId videos/{name}/qa_demo.png --frame 2976
npx remotion still src/remotion/index.ts CompositionId videos/{name}/qa_summary.png --frame 4500
npx remotion still src/remotion/index.ts CompositionId videos/{name}/qa_outro.png --frame 5800

Visual Inspection

Claude reads each still image (multimodal) and checks for:

QA Report

Claude generates a brief report:

=== Visual QA ===
✓ hero: Large title centered, good contrast
✓ features: 3 cards with distinct colors, well-spaced
⚠ demo: Content only fills ~60% width — consider wider layout
✓ summary: Clean layout, text readable
✓ outro: Triple-click animation visible

Result: 4/5 PASS, 1 WARNING

On FAIL: Auto-fix if possible (e.g., increase font size, adjust layout), re-render still, re-check. On WARN: Note in report, proceed (user can address during preview review). All PASS: Proceed to Step 10 silently.

Cleanup

After QA passes, remove stills:

rm -f videos/{name}/qa_*.png

Step 10: Render Video

Preview Render — The Only Mandatory Stop (Auto mode)

Auto mode: Render a 720p preview (~10x faster than 4K), then stop and ask user to confirm:

npx remotion render src/remotion/index.ts CompositionId videos/{name}/preview.mp4 --scale 0.33 --crf 28

# Report preview info
DUR=$(ffprobe -v quiet -show_entries format=duration -of csv=p=0 videos/{name}/preview.mp4 | cut -d. -f1)
SIZE=$(ls -lh videos/{name}/preview.mp4 | awk '{print $5}')
echo "Preview: ${DUR}s, ${SIZE}"

Ask user: "720p 预览已生成：videos/{name}/preview.mp4（时长 Xs）。请查看后确认："

• 确认，渲染4K → proceed to 4K render
• 需要修改 → user describes changes, Claude applies, re-render preview, repeat

This is the only stop in auto mode. Everything before this runs without prompting.

Interactive mode: Studio preview already done in Step 9. Skip preview render, go directly to 4K.

4K Render

npx remotion render src/remotion/index.ts CompositionId videos/{name}/output.mp4 --video-bitrate 16M

验证 4K:

ffprobe -v quiet -show_entries stream=width,height -of csv=p=0 videos/{name}/output.mp4
# 期望: 3840,2160

Optional: Render Vertical Highlight Clip (9:16)

npx remotion render src/remotion/index.ts MyVideoVertical videos/{name}/output_vertical.mp4 --video-bitrate 16M
npx remotion still src/remotion/index.ts Thumbnail9x16 videos/{name}/thumbnail_remotion_9x16.png

The vertical composition reuses Video.tsx with orientation: "vertical". All components auto-adapt for 9:16.

Step 11: Mix with Background Music

BGM Selection

Auto mode: Select BGM based on topic type:

• Tech/coding/tutorial → snow-stevekaldes-piano-397491.mp3 (calm)
• Product review/news/upbeat → perfect-beauty-191271.mp3 (positive)
• User provided custom BGM → use their file directly

Interactive mode: Ask user to choose from available tracks or provide their own.

If user provided a custom BGM file path, copy that instead:

# Default: auto-selected track
cp ~/.claude/skills/video-podcast-maker/assets/{selected-track}.mp3 videos/{name}/bgm.mp3

# Or user's custom BGM
cp /path/to/user-bgm.mp3 videos/{name}/bgm.mp3

Mix

ffmpeg -y \
  -i videos/{name}/output.mp4 \
  -stream_loop -1 -i videos/{name}/bgm.mp3 \
  -filter_complex "[0:a]volume=1.0[a1];[1:a]volume=0.05[a2];[a1][a2]amix=inputs=2:duration=first[aout]" \
  -map 0:v -map "[aout]" \
  -c:v copy -c:a aac -b:a 192k \
  videos/{name}/video_with_bgm.mp4

More BGM options, custom BGM, and volume tuning: See TROUBLESHOOTING.md.

Step 12: Add Subtitles

Auto mode: Skip subtitles — copy video_with_bgm.mp4 as final_video.mp4. Interactive mode: Ask user: "需要烧录字幕吗？字幕可以提高视频的可访问性。"

If user requested subtitles (any mode):

ffmpeg -y -i videos/{name}/video_with_bgm.mp4 \
  -vf "subtitles=videos/{name}/podcast_audio.srt:force_style='FontName=PingFang SC,FontSize=14,PrimaryColour=&H00333333,OutlineColour=&H00FFFFFF,Bold=1,Outline=2,Shadow=0,MarginV=20'" \
  -c:v libx264 -crf 18 -preset slow -s 3840x2160 \
  -c:a copy videos/{name}/final_video.mp4

If skipping subtitles:

cp videos/{name}/video_with_bgm.mp4 videos/{name}/final_video.mp4

Step 13: Complete Publish Info (Part 2)

从 timing.json 生成 B站章节：

00:00 开场
00:23 功能介绍
00:55 演示
01:20 总结

格式：MM:SS 章节标题，每段间隔 ≥5秒。

Step 14: Verify Output & Cleanup

14.1 Verification

VIDEO_DIR="videos/{name}"
echo "=== 文件检查 ==="
for f in podcast.txt podcast_audio.wav podcast_audio.srt timing.json output.mp4 final_video.mp4; do
  [ -f "$VIDEO_DIR/$f" ] && echo "✓ $f" || echo "✗ $f 缺失"
done

echo "=== 技术指标 ==="
RES=$(ffprobe -v quiet -select_streams v:0 -show_entries stream=width,height -of csv=p=0 "$VIDEO_DIR/final_video.mp4")
[ "$RES" = "3840,2160" ] && echo "✓ 分辨率: 3840x2160 (4K)" || echo "✗ 分辨率: $RES (非4K)"
DUR=$(ffprobe -v quiet -show_entries format=duration -of csv=p=0 "$VIDEO_DIR/final_video.mp4" | cut -d. -f1)
echo "✓ 时长: ${DUR}s"
SIZE=$(ls -lh "$VIDEO_DIR/final_video.mp4" | awk '{print $5}')
echo "✓ 文件大小: $SIZE"

14.2 Cleanup

Auto mode: Auto-clean temp files, report what was removed. Interactive mode: List files and ask for confirmation before deleting.

VIDEO_DIR="videos/{name}"
rm -f "$VIDEO_DIR"/part_*.wav "$VIDEO_DIR"/concat_list.txt
rm -f "$VIDEO_DIR"/output.mp4 "$VIDEO_DIR"/video_with_bgm.mp4
rm -f public/podcast_audio.wav public/timing.json public/media_manifest.json
rm -rf public/media/{name}
echo "✓ 临时文件已清理"

14.3 Final Report

=== 视频制作完成 ===
✓ 文件: final_video.mp4
✓ 分辨率: 3840x2160 (4K)
✓ 时长: XXs
✓ 大小: XXX MB
✓ 封面: thumbnail_remotion_16x9.png, thumbnail_remotion_4x3.png
✓ 发布信息: publish_info.md
✓ 临时文件已清理

Troubleshooting & Preferences

Full reference: Read TROUBLESHOOTING.md when encountering errors, when user asks about preferences or BGM options.