bilibili-summarizer

name	bilibili-summarizer
description	Use when the user shares a Bilibili video link and asks to summarize or save notes about the video.
allowed-tools	Bash, Write

Bilibili Video Summarizer

Fetch a Bilibili video's content via subtitles, Bilibili's AI summary API, or audio transcription — then save a structured note.

Configuration

Edit before first use:

SAVE_PATH=     # absolute path to the folder where notes will be saved
COOKIES_PATH=  # absolute path to your Bilibili cookies file (Netscape format)

Workflow

Step 1 — Get metadata

yt-dlp --dump-json --no-download "<url>"

Extract: title, uploader, duration, upload date, description, chapters, BV ID.

CID (needed for Step 2B) can be fetched without authentication:

python -c "
import urllib.request, json
bvid = '<BV_ID>'
r = urllib.request.urlopen('https://api.bilibili.com/x/web-interface/view?bvid=' + bvid)
d = json.loads(r.read())
print('cid:', d['data']['cid'], 'mid:', d['data']['owner']['mid'])
"

---

Step 2 — Fetch content (try in order)

Option A: yt-dlp + cookies file — subtitles (best)

Set COOKIES_PATH to wherever you save your Bilibili cookies file (see Configuration above).

Export cookies from your browser using an extension such as Get cookies.txt LOCALLY while on any Bilibili page, and save the file to COOKIES_PATH.

yt-dlp --cookies "<COOKIES_PATH>" \
  --skip-download --write-subs --write-auto-subs \
  --sub-lang "zh-Hans,zh-CN,zh,ai-zh,ai-zh-Hans,en" \
  --convert-subs srt \
  -o "/tmp/bili_%(id)s" \
  "<url>"

If a .srt file is created (e.g. /tmp/bili_<id>.zh-Hans.srt), read it and proceed to Step 3.

If the cookies file is missing or expired, prompt the user to re-export it.

---

Option B: Bilibili AI summary API (if Option A yields no subtitles)

Bilibili provides AI-generated summaries for many videos via an API:

python -c "
import json, urllib.request, http.cookiejar

# Load cookies from COOKIES_PATH
cj = http.cookiejar.MozillaCookieJar('<COOKIES_PATH>')
try:
    cj.load(ignore_discard=True, ignore_expires=True)
except:
    pass
cookie_str = '; '.join(f'{c.name}={c.value}' for c in cj if 'bilibili' in c.domain)

# Call AI summary endpoint
bvid = '<BV_ID>'
cid  = '<CID>'
req = urllib.request.Request(
    f'https://api.bilibili.com/x/web-interface/view/conclusion/get?bvid={bvid}&cid={cid}',
    headers={'Cookie': cookie_str, 'Referer': 'https://www.bilibili.com/', 'User-Agent': 'Mozilla/5.0'}
)
resp = json.loads(urllib.request.urlopen(req).read())
print(json.dumps(resp, ensure_ascii=False, indent=2))
"

If data.model_result is non-empty, extract summary and outline, then proceed to Step 3.

---

Option C: Download audio + Whisper transcription (if A and B both fail)

Inform the user that transcription takes 10–20 minutes, then run in background using a local pipeline (e.g. yt-dlp + Whisper). Proceed to Step 3 once the transcript is ready.

---

Option D: Description + chapters only (last resort)

Use only when all above options fail. Always add this notice to the note:

> ⚠️ This summary is based on the video description and chapter titles only — no transcript was available.

---

Step 3 — Summarize

From the content obtained (subtitles / AI summary / transcript):

• 3–5 sentence overview of the main topic
• 5–10 key points as a bullet list
• Detailed notes organised by chapter or theme

---

Step 4 — Write the note

Save to <SAVE_PATH>/<video title>.md:

# [Video Title]

> **Uploader**: [name] | **Duration**: [hh:mm:ss] | **Published**: [date]

## Summary

[3–5 sentences]

## Key Points

- [point 1]
- [point 2]

## Detailed Notes

[Expanded notes by chapter or theme — should be substantive when a transcript is available]

## Source

- [Video Title](<Bilibili URL>)

---

Notes

• Strip illegal filename characters: \ / : * ? " < > |
• If chapters exist, use them as section headings in Detailed Notes
• Option A requires a valid cookies file — re-export if subtitles fail unexpectedly
• Option C is slow — always run in background and inform the user