SHILIU · 创作资源
视频播客转写与结构化摘要 · 创作技能
面向视频与播客内容研究：先判断平台与媒体类型，优先提取字幕，无字幕则下载音频、按 25MB 上限分段，调用 Groq Whisper 转写，再按视频或播客模板输出概览、要点、金句与行动项。含错误处理与依赖说明。
这个技能解决什么问题
面向视频与播客内容研究：先判断平台与媒体类型，优先提取字幕，无字幕则下载音频、按 25MB 上限分段，调用 Groq Whisper 转写，再按视频或播客模板输出概览、要点、金句与行动项。含错误处理与依赖说明。
适用任务
把用户提供的视频或播客链接转成完整转写文本，并生成结构化摘要。
需要的输入
一个受支持的媒体 URL（YouTube、Bilibili、X 视频、小宇宙、Apple Podcasts 或直链音视频）；本机安装 yt-dlp、ffmpeg、curl；无字幕内容需用户自备 GROQ_API_KEY；必要时提供浏览器 Cookie 以应对平台风控。
交付内容
转写文本（字幕 VTT 内容或 Whisper 返回文本）与结构化摘要：视频含概览、3-5 条要点、金句、行动项；播客含概览、分章摘要、5-8 条要点、金句、行动项。
任务示例
把这个 B 站视频 https://www.bilibili.com/video/BV1xxxxx 转成文字，并给我一份带要点和金句的摘要。
输出示例
示意：先输出「转写文本」段落（来自字幕或 Whisper，标注来源），再输出「摘要」：概览 1-2 句、要点 3-5 条、金句若干、行动项（如有）。内容以实际转写为准，此处不展示真实结果。
来源与使用
作者 iBigQiang；许可 MIT。已核对来源与包结构，未据此宣称完成运行实测。
机器可读详情 · 读取 SKILL.md · 下载完整技能包
技能说明全文
---
name: video
description: Video & Podcast Digest — send a video/podcast link, get full transcript + structured summary. Supports YouTube, Bilibili, X/Twitter video, Xiaoyuzhou, Apple Podcasts, and direct audio/video links. Uses yt-dlp for subtitles and Groq Whisper for transcription.
---
# Video & Podcast Digest Skill
> Send a video/podcast link → get full transcript + structured summary
## Supported Platforms
| Platform | Type | Subtitles | Whisper Transcription |
|----------|------|-----------|----------------------|
| YouTube | Video | ✅ | ✅ |
| Bilibili | Video | ✅ | ✅ |
| X/Twitter | Video | ❌ | ✅ |
| Xiaoyuzhou (小宇宙) | Podcast | ❌ | ✅ |
| Apple Podcasts | Podcast | ❌ | ✅ |
| Direct links (mp3/mp4/m3u8) | Any | ❌ | ✅ |
## Trigger
Auto-triggered when a media URL is detected:
- YouTube: `youtube.com`, `youtu.be`
- Bilibili: `bilibili.com`, `b23.tv`
- X/Twitter: `x.com`, `twitter.com` (tweets with video)
- Xiaoyuzhou: `xiaoyuzhoufm.com`
- Apple Podcasts: `podcasts.apple.com`
- Direct: `.mp3`, `.mp4`, `.m3u8`, `.m4a`, `.webm`
## Pipeline
### Step 0: Detect Media Type
| URL Pattern | Type | Pipeline |
|-------------|------|----------|
| `xiaoyuzhoufm.com/episode/` | Podcast | → Step 1b (Xiaoyuzhou) |
| `podcasts.apple.com` | Podcast | → Step 1c (Apple) |
| `bilibili.com`, `b23.tv` | Video | → Step 1d (Bilibili API) |
| `.mp3`, `.m4a` direct link | Audio | → Step 2b (direct download) |
| Other | Video | → Step 1a (subtitle extraction) |
### Step 1a: Video — Extract Subtitles
```bash
rm -f /tmp/media_sub*.vtt /tmp/media_audio.mp3 /tmp/media_transcript*.json /tmp/media_segment_*.mp3 2>/dev/null || true
# YouTube (prefer English, fallback Chinese)
yt-dlp --skip-download --write-auto-sub --sub-lang "en,zh-Hans" -o "/tmp/media_sub" "VIDEO_URL"
# Bilibili
yt-dlp --skip-download --write-auto-sub --sub-lang "zh-Hans,zh" -o "/tmp/media_sub" "VIDEO_URL"
```
Check for subtitles:
```bash
ls /tmp/media_sub*.vtt 2>/dev/null
```
- **Has subtitles** → Read VTT content, skip to Step 3
- **No subtitles** → Step 2a (download audio)
### Step 1b: Xiaoyuzhou (小宇宙) — Extract Audio URL
```bash
AUDIO_URL=$(curl -sL -H "User-Agent: Mozilla/5.0 (Macintosh; Intel Mac OS X 10_15_7) AppleWebKit/537.36" \
"EPISODE_URL" \
| grep -oE 'https://media\.xyzcdn\.net/[^"]+\.(m4a|mp3)' \
| head -1)
echo "Audio URL: $AUDIO_URL"
curl -L -o /tmp/media_audio.mp3 "$AUDIO_URL"
```
→ Step 2b
### Step 1c: Apple Podcasts — via yt-dlp
```bash
yt-dlp -f "ba[ext=m4a]/ba/b" --extract-audio --audio-format mp3 --audio-quality 5 \
-o "/tmp/media_audio.%(ext)s" "APPLE_PODCAST_URL"
```
→ Step 2b
### Step 1d: Bilibili — API Direct Audio Stream
```bash
BV="BV1xxxxx"
curl -s "https://api.bilibili.com/x/web-interface/view?bvid=$BV" \
-H "User-Agent: Mozilla/5.0" -H "Referer: https://www.bilibili.com/" \
| python3 -c "import json,sys; d=json.load(sys.stdin)['data']; print(f\"Title: {d['title']}\nDuration: {d['duration']}s\nCID: {d['cid']}\")"
CID=<CID from previous step>
AUDIO_URL=$(curl -s "https://api.bilibili.com/x/player/playurl?bvid=$BV&cid=$CID&fnval=16&qn=64" \
-H "User-Agent: Mozilla/5.0" -H "Referer: https://www.bilibili.com/" \
| python3 -c "import json,sys; print(json.load(sys.stdin)['data']['dash']['audio'][0]['baseUrl'])")
curl -L -o /tmp/media_audio.m4s \
-H "User-Agent: Mozilla/5.0" -H "Referer: https://www.bilibili.com/" "$AUDIO_URL"
ffmpeg -y -i /tmp/media_audio.m4s -acodec libmp3lame -q:a 5 /tmp/media_audio.mp3
```
→ Step 2b
### Step 2a: Video — Download Audio (when no subtitles)
```bash
yt-dlp --cookies-from-browser chrome -f "ba[ext=m4a]/ba/b" --extract-audio --audio-format mp3 --audio-quality 5 \
-o "/tmp/media_audio.%(ext)s" "VIDEO_URL"
```
### Step 2b: Check Audio Size & Segment
```bash
FILE_SIZE=$(stat -f%z /tmp/media_audio.* 2>/dev/null || stat -c%s /tmp/media_audio.* 2>/dev/null)
echo "File size: $FILE_SIZE bytes"
```
- **≤ 25MB** → Step 2c (transcribe directly)
- **> 25MB** → Split into 10-minute segments:
```bash
DURATION=$(ffprobe -v error -show_entries format=duration -of csv=p=0 /tmp/media_audio.* | head -1)
SEGMENT_SEC=600
SEGMENTS=$(python3 -c "import math; print(math.ceil(float('$DURATION')/$SEGMENT_SEC))")
for i in $(seq 0 $((SEGMENTS-1))); do
START=$((i * SEGMENT_SEC))
ffmpeg -y -i /tmp/media_audio.* -ss $START -t $SEGMENT_SEC -acodec libmp3lame -q:a 5 \
"/tmp/media_segment_${i}.mp3" 2>/dev/null
done
```
→ Call Step 2c for each segment **sequentially** (parallel triggers Groq 524 timeout)
### Step 2c: Whisper Transcription
**Prerequisite**: `GROQ_API_KEY` environment variable
```bash
if [ -z "$GROQ_API_KEY" ]; then
echo "❌ GROQ_API_KEY not set. Get one at: https://console.groq.com/keys"
exit 1
fi
curl -s -X POST "https://api.groq.com/openai/v1/audio/transcriptions" \
-H "Authorization: Bearer $GROQ_API_KEY" \
-H "Content-Type: multipart/form-data" \
-F "file=@AUDIO_FILE" \
-F "model=whisper-large-v3-turbo" \
-F "response_format=verbose_json" \
-F "language=zh" \
> /tmp/media_transcript.json
python3 -c "import json; print(json.load(open('/tmp/media_transcript.json'))['text'])"
```
### Step 3: Structured Summary
**Video (≤20 min)**:
1. **Overview** (1-2 sentences)
2. **Key Points** (3-5 bullet points)
3. **Notable Quotes** (if any)
4. **Action Items** (if applicable)
**Podcast (>20 min)**:
1. **Overview** (2-3 sentences)
2. **Chapter Summary** (segmented by topic)
3. **Key Points** (5-8 bullet points)
4. **Notable Quotes**
5. **Action Items** (if applicable)
## Error Handling
| Situation | Action |
|-----------|--------|
| No subtitles + no GROQ_API_KEY | Prompt user to set API key |
| Audio >25MB | ffmpeg segment (10min/segment), transcribe sequentially |
| Podcast >2 hours | Warn user, confirm before proceeding |
| Groq 524 timeout | Do NOT parallelize — transcribe sequentially, sleep 5-8s between |
| Groq 429 rate limit | Wait for retry-after header, then retry |
| yt-dlp Bilibili 412 | Use Bilibili API instead (Step 1d) |
| yt-dlp YouTube bot detection | Add `--cookies-from-browser chrome` |
| Spotify links | Not supported (DRM protected) |
## Dependencies
- `yt-dlp`: video download + subtitle extraction
- `ffmpeg`: audio conversion + segmentation
- `curl`: audio download, API calls
- `GROQ_API_KEY`: Whisper transcription (free at https://console.groq.com/keys)
技能是任务材料，不会自动授予账户权限、付费授权或其他系统权限。请让 Agent 按说明核对依赖，再在授权范围内使用。