完整的抖音（抖音/TikTok中国）视频转录解决方案。提取音频、转录为文本并生成结构化摘要。
版本历史
| 版本 | 变更 |
|---|
| 2.0.0 | 模块化架构、浏览器DOM提取、DASH支持、Docker Whisper、结构化输出格式、扩展故障排除指南 |
| 1.0.0 | 初始发布、基本转录 |
架构
...
（
注意：由于字符限制，完整的cn_skill_md_content未全部包含，请根据需要自行补充）
A complete solution for transcribing Douyin (抖音/TikTok China) videos. Extracts audio, transcribes speech to text, and generates structured summaries.
Version History
| Version | Changes |
|---|
| 2.0.0 | Modular architecture, improved workflow, browser DOM extraction |
| 1.0.0 | Initial release, basic transcription |
Architecture
\\\
User Input (Douyin Link/File)
│
▼
┌─────────────────────────────────────────┐
│ Workflow Orchestrator │
├─────────────────────────────────────────┤
│ Step 1: Fetcher → Get video file │
│ Step 2: Transcriber → Extract & convert│
│ Step 3: Analyzer → Structure output │
│ Step 4: Output → Save results │
└─────────────────────────────────────────┘
\\\
Core Features
- Video Fetching: Browser-based DOM extraction for CDN URLs
- Audio Extraction: ffmpeg-powered audio conversion
- Speech-to-Text: Whisper ASR with multiple model options
- Content Analysis: Auto-structured transcripts with key points
- Multi-format Support: Video links, local files, image notes
Prerequisites
| Tool | Purpose | Install |
|---|
| curl | Download files | Built-in (Windows: \curl.exe\) |
| ffmpeg | Audio extraction/merge | \winget install Gyan.FFmpeg\ |
| Whisper | Transcription | \pip install openai-whisper\ or Docker |
| Browser | Video extraction | OpenClaw profile required |
Docker Whisper (Recommended):
\\\ash
docker run -d -p 9000:9000 --name whisper-asr onerahmet/openai-whisper-asr-webservice:latest
\\\
Workflow
Step 0: Input Classification
| Input Type | Detection | Action |
|---|
| Video link (\/video/\) | URL pattern | Full workflow |
| Image note (\/note/\) | URL pattern | Snapshot only |
| Local video file | File path | Start from Step 2 |
| Text input | Plain text | Start from Step 3 |
Step 1: Fetch Video
1.1 Resolve Short URL
\\\ash
# Windows PowerShell
curl.exe -sL -o NUL -w "%{url_effective}" "https://v.douyin.com/xxx/"
# macOS/Linux
curl -sL -o /dev/null -w '%{url_effective}' "https://v.douyin.com/xxx/"
\\\
Output: \https://www.douyin.com/video/7616020798351871284\
1.2 Open Video Page
\\\
browser(action='open', profile='openclaw', url='https://www.douyin.com/video/{VIDEO_ID}')
\\\
Wait 10-15 seconds for page to load completely.
1.3 Extract Video URL (Browser DOM Method)
\\\javascript
browser(action='act', targetId='PAGE_ID', request={
"kind": "evaluate",
"fn": "(() => {
const entries = performance.getEntriesByType('resource');
const videoEntries = entries.filter(e => {
const name = e.name.toLowerCase();
return name.includes('douyinvod') &&
(name.includes('.mp4') || name.includes('video'));
});
if (videoEntries.length > 0) {
const video = videoEntries[videoEntries.length - 1];
return {
url: video.name,
type: video.name.includes('.mp4') ? 'mp4' : 'dash'
};
}
return null;
})()"
})
\\\
Important Notes:
- \ct\ action requires nested \
equest\ object with \kind\ and \ n\
- Wrong: \rowser(action='act', fn='...')\
- Correct: \rowser(action='act', request={"kind": "evaluate", "fn": "..."})\
1.4 Download Video
\\\ash
curl.exe -L -H "Referer: https://www.douyin.com/" -o video.mp4 ""
\\\
Referer header is required, otherwise 403.
Step 2: Transcribe Audio
2.1 Extract Audio
\\\ash
# For MP4 videos
ffmpeg -i video.mp4 -ar 16000 -ac 1 -c:a pcm_s16le audio.wav -y
# For DASH videos (need merge)
ffmpeg -i video.mp4 -i audio.mp4 -c copy merged.mp4 -y
ffmpeg -i merged.mp4 -ar 16000 -ac 1 -c:a pcm_s16le audio.wav -y
\\\
Parameters:
- \-ar 16000\: 16kHz sample rate (Whisper requirement)
- \-ac 1\: Mono channel
- \-c:a pcm_s16le\: 16-bit PCM
2.2 Transcribe with Docker Whisper
\\\ash
curl.exe -X POST "http://localhost:PORT/asr" -F "audio_file=@audio.wav"
\\\
2.3 Alternative: Local Whisper
\\\ash
python -m whisper audio.wav --model small --language zh
\\\
Model Selection:
| Model | Size | 5-min Video (CPU) | Accuracy | Use Case |
|---|
| tiny | 75MB | ~30s | Fair | Quick preview |
| base | 142MB | ~1min | Good | Daily use |
| small | 466MB | ~3min | Better | Recommended |
| medium | 1.5GB | ~8min | Best | High accuracy |
Step 3: Analyze Content
Agent processes transcript and generates:
- Correct homophones
- Fix speaker names
- Remove filler words
- Add paragraph breaks
- Create sections
- Main ideas
- Important quotes
- 3-5 topic tags
Step 4: Save Output
Transcript Format
\\\markdown
# {Title}
作者: {Author}
来源: 抖音
日期: {Date}
转录时间: {Transcription Date}
摘要
{Summary}
正文
{Transcript content with paragraphs}
要点
- {Key point 1}
- {Key point 2}
- {Key point 3}
标签
#{tag1} #{tag2} #{tag3}
\\\
File Naming Convention
\\\
{VIDEO_ID}-抖音转录.md
\\\
Troubleshooting
| Stage | Issue | Solution |
|---|
| Step 1 | Short URL fails | Check link completeness, remove share text |
| Step 1 | JS returns null | Wait 15-20s and retry, increase timeout |
| Step 1 | Download 403 | URL expired, re-fetch from browser |
| Step 1 | DASH no audio | Merge with \ fmpeg -i video -i audio -c copy\ |
| Step 2 | ffmpeg not installed | \winget install Gyan.FFmpeg\ |
| Step 2 | Whisper service down | \docker start whisper-asr\ |
| Step 2 | Transcription slow | 10-min video takes 15-20 min on CPU |
| Step 2 | Poor quality | Use larger model (medium) |
Image Note Handling
Image notes (\/note/\) don't need transcription:
\\\
- browser(action='open', profile='openclaw', url='IMAGE_NOTE_URL')
- browser(action='snapshot')
- Extract content from snapshot
- Save to output directory
\\\
Edge Cases
- Article links (\/article/\): Use browser snapshot, no transcription
- Douyin AI summary: Extract from page as supplement
- Other platforms: Use yt-dlp for YouTube/Bilibili
- Live streams: Not supported
Related Modules
This skill can be extended with standalone modules:
| Module | Purpose |
|---|
| douyin-fetcher | Video fetching only |
| douyin-transcriber | Audio transcription only |
| douyin-analyzer | Content analysis only |
| douyin-orchestrator | Workflow coordination |
License
MIT-0 License - Free to use, modify, and redistribute.
by