Openakita/skills@bilibili Watcher

Bilibili & YouTube Watcher — 双平台字幕提取

从 Bilibili（B站）和 YouTube 视频中提取字幕/转录文本，支持多语言、会员视频和内容问答。

When to Use This Skill

• User shares a Bilibili link and wants subtitles or a summary
• User shares a YouTube link and wants transcript extraction via yt-dlp
• User needs subtitles from member-only (大会员) Bilibili videos
• User wants to search or query content within a video's transcript
• User wants to compare subtitles across languages
• User needs to extract subtitles from videos with hardcoded subs (OCR not included — only soft subs)
• User wants batch subtitle extraction from a playlist or series

Prerequisites

Install yt-dlp

yt-dlp is a feature-rich command-line audio/video downloader that also extracts subtitles.

Via pip (recommended):

pip install yt-dlp

Via package manager:

# macOS
brew install yt-dlp

# Ubuntu/Debian
sudo apt install yt-dlp

# Windows (scoop)
scoop install yt-dlp

Verify installation:

yt-dlp --version

Optional: ffmpeg

Some subtitle formats require ffmpeg for conversion:

# macOS
brew install ffmpeg

# Ubuntu/Debian
sudo apt install ffmpeg

# Windows (scoop)
scoop install ffmpeg

Cookie Setup for Bilibili Member Videos

Bilibili member-only (大会员) content requires authentication cookies.

Method 1: Export cookies from browser

Install a browser extension like "Get cookies.txt LOCALLY" and export cookies for bilibili.com:

# Use the exported cookies file
yt-dlp --cookies cookies.txt "https://www.bilibili.com/video/BV..."

Method 2: Use browser cookies directly

# yt-dlp can read cookies from your browser
yt-dlp --cookies-from-browser chrome "https://www.bilibili.com/video/BV..."
yt-dlp --cookies-from-browser firefox "https://www.bilibili.com/video/BV..."
yt-dlp --cookies-from-browser edge "https://www.bilibili.com/video/BV..."

Security note: Cookie files contain your login session. Do not share them or commit them to version control.

Instructions

Step 1: Identify the Platform and URL Format

Bilibili URL Formats

YouTube URL Formats

Step 2: Extract Subtitles

Bilibili Subtitle Extraction

List available subtitles:

yt-dlp --list-subs "https://www.bilibili.com/video/BV..."

Download subtitles only (no video):

# Download all available subtitles
yt-dlp --write-sub --skip-download "https://www.bilibili.com/video/BV..."

# Download auto-generated subtitles as well
yt-dlp --write-sub --write-auto-sub --skip-download "https://www.bilibili.com/video/BV..."

# Download specific language (Chinese)
yt-dlp --write-sub --sub-lang zh-CN --skip-download "https://www.bilibili.com/video/BV..."

# Convert to SRT format
yt-dlp --write-sub --sub-lang zh-CN --convert-subs srt --skip-download "https://www.bilibili.com/video/BV..."

Member-only videos (require cookies):

yt-dlp --cookies-from-browser chrome --write-sub --skip-download "https://www.bilibili.com/video/BV..."

YouTube Subtitle Extraction

List available subtitles:

yt-dlp --list-subs "https://www.youtube.com/watch?v=VIDEO_ID"

Download subtitles:

# English subtitles
yt-dlp --write-sub --sub-lang en --skip-download "https://www.youtube.com/watch?v=VIDEO_ID"

# Auto-generated subtitles
yt-dlp --write-auto-sub --sub-lang en --skip-download "https://www.youtube.com/watch?v=VIDEO_ID"

# Multiple languages
yt-dlp --write-sub --sub-lang "en,zh-Hans,ja" --skip-download "URL"

# Convert to plain text (SRT format)
yt-dlp --write-auto-sub --sub-lang en --convert-subs srt --skip-download "URL"

Step 3: Parse Subtitle Files

yt-dlp downloads subtitles in various formats. Here's how to parse common ones:

import re
import json

def parse_srt(filepath: str) -> list[dict]:
    """Parse SRT subtitle file into structured segments."""
    with open(filepath, 'r', encoding='utf-8') as f:
        content = f.read()

    segments = []
    blocks = content.strip().split('\n\n')

    for block in blocks:
        lines = block.strip().split('\n')
        if len(lines) >= 3:
            time_match = re.match(
                r'(\d{2}):(\d{2}):(\d{2}),(\d{3}) --> (\d{2}):(\d{2}):(\d{2}),(\d{3})',
                lines[1]
            )
            if time_match:
                h, m, s = int(time_match[1]), int(time_match[2]), int(time_match[3])
                start_sec = h * 3600 + m * 60 + s
                text = ' '.join(lines[2:]).strip()
                # Remove HTML tags from auto-generated subs
                text = re.sub(r'<[^>]+>', '', text)
                if text:
                    segments.append({
                        'start': start_sec,
                        'text': text
                    })

    return segments


def parse_json3(filepath: str) -> list[dict]:
    """Parse YouTube JSON3 subtitle format."""
    with open(filepath, 'r', encoding='utf-8') as f:
        data = json.load(f)

    segments = []
    for event in data.get('events', []):
        start_ms = event.get('tStartMs', 0)
        segs = event.get('segs', [])
        text = ''.join(s.get('utf8', '') for s in segs).strip()
        if text and text != '\n':
            segments.append({
                'start': start_ms / 1000,
                'text': text
            })

    return segments


def segments_to_text(segments: list[dict]) -> str:
    """Convert segments to plain text with timestamps."""
    lines = []
    for seg in segments:
        minutes = int(seg['start'] // 60)
        seconds = int(seg['start'] % 60)
        lines.append(f"[{minutes:02d}:{seconds:02d}] {seg['text']}")
    return '\n'.join(lines)

Step 4: Summarize or Query the Content

Once you have the transcript text, generate summaries or answer questions about the content.

For summarization, combine all subtitle text and apply a structured prompt:

Based on the following video transcript, provide:

1. **概要** (Executive Summary): 2-3 sentences in the video's language
2. **要点** (Key Points): Bulleted list with timestamps [MM:SS]
3. **详细笔记** (Detailed Notes): Organized by topic sections
4. **问答** (Q&A): Answer any specific questions the user has

Transcript:
{full_transcript_text}

For Q&A, search the transcript for relevant segments first, then answer based on context.

Workflows

Workflow 1: Quick Bilibili Subtitle Extraction

User says: "提取这个B站视频的字幕: https://www.bilibili.com/video/BV..."

1. Run yt-dlp --list-subs to check available subtitles
2. Download Chinese subtitles: yt-dlp --write-sub --sub-lang zh-CN --convert-subs srt --skip-download URL
3. Parse the SRT file
4. Present the clean transcript to the user

Workflow 2: Bilibili Member Video

User says: "这是大会员视频，帮我提取字幕"

1. Inform user that cookies are needed
2. Use --cookies-from-browser chrome (or user's preferred browser)
3. Extract subtitles with authentication
4. If cookies fail, guide user to export cookies.txt manually

Workflow 3: YouTube Multi-Language

User says: "Get both English and Chinese subtitles from this YouTube video"

1. List available subtitle languages
2. Download both en and zh-Hans subtitles
3. Parse both files
4. Present side-by-side or merged view

Workflow 4: Video Content Q&A

User says: "视频里有没有提到关于 X 的内容?"

1. Extract full transcript
2. Search for keywords related to X
3. Return matching segments with timestamps
4. Provide a concise answer based on the matching content

Workflow 5: Batch Playlist Extraction

User provides a playlist or series URL:

1. Use yt-dlp --flat-playlist to list all videos
2. Extract subtitles from each video sequentially
3. Save each transcript as a separate file
4. Generate a combined index with video titles and file paths

Workflow 6: Bilibili Bangumi (番剧) Subtitles

User shares a bangumi URL:

1. Bangumi often has official multi-language subtitles
2. Use --list-subs to show all available languages
3. Download preferred language(s)
4. Note: Some bangumi require 大会员 cookies

Output Format

Transcript Output

# 📝 Subtitles: [Video Title]

**Platform**: Bilibili / YouTube
**Language**: 中文 (zh-CN)
**Duration**: ~XX minutes
**Subtitle Type**: Manual / Auto-generated

---

[00:00] 大家好，欢迎来到今天的视频
[00:05] 今天我们要讨论的话题是...
[00:12] 首先我们来看一下背景
...

Summary Output

# 📋 Video Summary: [Title]

## 概要
[2-3 sentence summary in the video's language]

## 要点
- **[00:00]** 开场介绍和主题说明
- **[02:15]** 第一个核心观点
- **[08:30]** 关键论据和数据
- **[15:00]** 实际演示
- **[22:45]** 总结与下一步

## 详细笔记

### 第一部分: [主题] (00:00 - 05:30)
[详细内容笔记]

### 第二部分: [主题] (05:30 - 12:00)
[详细内容笔记]

Common Pitfalls

1. Bilibili Geo-Restrictions

Problem: Some Bilibili content is restricted to mainland China.

Solutions:

• Use a proxy or VPN with a Chinese IP: yt-dlp --proxy socks5://127.0.0.1:1080 URL
• Set the --geo-bypass flag: yt-dlp --geo-bypass URL
• For persistent issues, use --geo-bypass-country CN

yt-dlp --geo-bypass-country CN --write-sub --skip-download "URL"

2. Member-Only Content Without Cookies

Problem: yt-dlp returns an error or empty subtitles for 大会员 videos.

Solution: Always check if the video requires 大会员 access. If so, cookies are mandatory:

# If this fails:
yt-dlp --list-subs "URL"
# ERROR: This video requires premium membership

# Try with cookies:
yt-dlp --cookies-from-browser chrome --list-subs "URL"

If browser cookie extraction fails (common on Linux), export cookies manually to a cookies.txt file.

3. No Subtitles Available

Problem: Many Bilibili videos, especially older ones or user-generated content, have no subtitles at all.

Solution: Inform the user clearly. Unlike YouTube, Bilibili does not always generate auto-subtitles. The video may only have hardcoded (burned-in) subtitles which require OCR — beyond the scope of this skill.

4. yt-dlp Version Issues

Problem: Bilibili frequently changes its API, causing older yt-dlp versions to fail.

Solution: Always ensure yt-dlp is up to date:

pip install -U yt-dlp
# or
yt-dlp -U

5. Subtitle Format Inconsistencies

Problem: Different videos return subtitles in different formats (SRT, VTT, JSON3, ASS).

Solution: Use --convert-subs srt to normalize all subtitles to SRT format:

yt-dlp --write-sub --convert-subs srt --skip-download "URL"

6. Rate Limiting on Bilibili

Problem: Rapid successive requests may get temporarily blocked.

Solutions:

• Add delays between batch requests: --sleep-requests 2
• Use --sleep-interval 5 for playlists
• Limit concurrent downloads: --max-downloads 10

7. Short Link Resolution

Problem: Bilibili short links (b23.tv) may not resolve properly.

Solution: yt-dlp handles most redirects automatically, but if it fails:

# Manually resolve the short link first
curl -sI "https://b23.tv/aBcDeFg" | grep -i location

Multi-Language Subtitle Support

Common Language Codes

Checking Available Languages

# Bilibili
yt-dlp --list-subs "https://www.bilibili.com/video/BV..."

# YouTube
yt-dlp --list-subs "https://www.youtube.com/watch?v=..."

The output shows both manual and auto-generated subtitle tracks with their language codes.

Platform Comparison