Bilibili Scraper — videos, comments, danmaku, UP主 stats and tags

Extract data from Bilibili (哔哩哔哩 / B站) videos, search results, comment threads and danmaku (弹幕). Use keywords or video URLs to scrape titles, uploaders, plays, danmaku, likes, coins, favorites, shares, exact comment counts, tags, descriptions and top comments. Export as JSON, CSV, Excel or XML, run it via API, schedule it, or connect it to any AI agent through the Apify MCP server.

Videos, stats, tags and search: no login, no cookies, no API key, no wbi signature to renew.

Danmaku (弹幕) need no login either — the timed bullet comments come back for free, with the exact second each one appears.

Comments are the exception. Bilibili hands out only 3 comments per video to anonymous requests — that is a platform rule, not a limit of this Actor. To get more, scan a QR code with the free Bilibili mobile app once; the session is remembered and every later run is hands-free. See How to log in.

What does Bilibili Scraper do?

• Search chosen keywords on Bilibili and collect every video result
• Scrape specific videos by URL or BVID — no search needed
• Scrape every video from an uploader (UP主) — works best with your own Bilibili cookie, see the FAQ
• Get full engagement stats per video: plays, danmaku, likes, coins, favorites, shares, comment count
• Collect top comments per video with author, text, likes and reply count
• Collect danmaku (弹幕) with the exact second, colour and mode — no login needed
• Pull tags and the full description from each video page
• Sort results by relevance, most played, newest, most danmaku or most favorited
• De-duplicate automatically across keywords and URL lists
• Export in multiple formats: JSON, CSV, Excel, XML
• Connect to any AI agent through the Apify MCP server, or call it from the Python and JavaScript SDKs, REST API or webhooks
• No login and no cookies for video, search and uploader data — everything runs from the browser session the Actor opens itself. Comments beyond 3 per video need a one-time scan.

What data can you get from Bilibili?

👤 Uploader name	🆔 Uploader ID	🔗 Channel URL
▶️ Plays	💬 Danmaku	❤️ Likes
🪙 Coins	⭐ Favorites	🔁 Shares
📝 Full description	🏷️ Tags	🖼️ Cover image URL
⏱️ Duration (text + seconds)	📅 Publish date + exact timestamp	🆔 BVID + numeric AID
💬 Total comment count	🗣️ Top comments with likes	🔍 Keyword that found it

How to scrape Bilibili with Bilibili Scraper?

Five steps, no coding:

1. Create a free Apify account with your email.
2. Open Bilibili Scraper.
3. Add the keywords you want to search, or paste video URLs.
4. Click Start and wait for the run to finish.
5. Download your data as JSON, CSV, Excel or XML from the Output tab.

Input

Two ways in — use either or both:

• keywords: search Bilibili and collect the video results
• videoUrls: scrape these exact videos by URL or BVID

{
  "keywords": ["ai", "原神"],
  "videoUrls": ["https://www.bilibili.com/video/BV1VZWc6dECG/"],
  "searchType": "video",
  "sortBy": "most_played",
  "maxItemsPerKeyword": 100,
  "maxPagesPerKeyword": 10,
  "includeStats": true,
  "includeComments": false,
  "maxCommentsPerVideo": 20,
  "commentSort": "hot"
}

Field	Type	Default	Description
keywords	array	—	Search terms. One run can take many
videoUrls	array	—	Full URLs or bare BVIDs
userUrls	array	—	Uploader space URLs or numeric IDs
bilibiliCookie	string	—	Optional. Paste a cookie only if you do not want to scan
loginTimeoutSeconds	integer	0	How long to keep a login QR available (0 = never ask; the code refreshes itself every ~3 min)
resetStoredSession	boolean	false	Delete the saved login and ask for a fresh QR
failWithoutLogin	boolean	false	If the scan never happens, stop instead of returning partial data
searchType	string	video	video or all
sortBy	string	relevance	relevance, most_played, newest, most_danmaku, most_favorited
maxItemsPerKeyword	integer	100	Videos per keyword (1–5000)
maxPagesPerKeyword	integer	10	Result pages per keyword (≈40 videos each)
includeStats	boolean	true	Attach likes, coins, favorites, shares, comment count, description, tags
includeComments	boolean	false	Also collect top-level comments (3 per video without a login)
maxCommentsPerVideo	integer	20	Comments per video (0–200). Anonymous Bilibili returns 3 no matter what this says — log in for more
includeDanmaku	boolean	false	Also collect danmaku (弹幕). No login required
maxDanmakuPerVideo	integer	200	Danmaku per video (0–5000). Each danmaku is one billed result row
commentSort	string	hot	hot (most liked) or new (newest)

At least one of keywords, videoUrls or userUrls is required.

Output

Results land in a dataset. Video, comment and danmaku rows share it — split them on record_type.

Danmaku row

{
  "record_type": "danmaku",
  "bvid": "BV1VZWc6dECG",
  "title": "【Re:从零开始的AI学习】第一季总集篇",
  "text": "不要 Fear Of Missing Out (FOMO)！",
  "time_in_video": 2754.324,
  "mode": 1,
  "font_size": 25,
  "color": "#ffffff",
  "sent_at": "2026-10-24T18:12:45+00:00",
  "dmid": "2225301461351363840"
}

You don't have to pick one or the other. Turn on includeComments and a single run returns the videos and their comments. Three ready-made table views are included:

View	Shows	Use it when
Overview	video rows only	you just want the videos
Comments	comment rows only	you just want the comments
Videos + comments (one table)	both, in one grid	you want to read a video next to what people said under it
Danmaku (弹幕)	danmaku rows only	you are studying reactions at a specific moment

Every comment row carries its parent video's title, play_count, comment_count and url, so it stands on its own in a spreadsheet, a pivot table or a database — no join needed. Only message is empty on video rows, because a video has no comment text.

Video row

{
  "record_type": "video",
  "bvid": "BV1VZWc6dECG",
  "aid": 116893619560671,
  "url": "https://www.bilibili.com/video/BV1VZWc6dECG/",
  "title": "【Re:从零开始的AI学习】第一季总集篇｜从入门到用好 AI",
  "author": "卡普迪姆",
  "author_mid": 312249633,
  "author_url": "https://space.bilibili.com/312249633",
  "cover_url": "https://i0.hdslb.com/bfs/archive/0bc4fd83ef1159aa3bc3454ec0a477e161191f.jpg",
  "play_count": 940040,
  "danmaku_count": 4316,
  "like_count": 72654,
  "coin_count": 21456,
  "favorite_count": 88120,
  "share_count": 7304,
  "comment_count": 1892,
  "duration_text": "54:57",
  "duration_seconds": 3297,
  "publish_date": "2026-10-13",
  "publish_datetime": "2026-10-13T00:57:21+00:00",
  "description": "从零开始学AI……",
  "tags": "AI开源先锋计划, 科普, AI, 人工智能, B 站大学, 大模型, Agent",
  "keyword": "ai",
  "scraped_at": "2026-10-25T05:02:11.883Z"
}

Comment row

{
  "record_type": "comment",
  "bvid": "BV1VZWc6dECG",
  "aid": 116893619560671,
  "comment_id": 2345678901,
  "parent_id": 0,
  "author": "老崔有话说",
  "author_mid": 12345678,
  "message": "讲得太清楚了，收藏了慢慢看",
  "like_count": 15,
  "reply_count": 3,
  "publish_date": "2026-10-24T14:23:56+00:00",
  "keyword": "ai",
  "scraped_at": "2026-10-25T05:02:11.883Z"
}

How much does it cost to scrape Bilibili?

Pay per result — only for rows actually written to the dataset. A run that finds nothing costs you almost nothing.

• $5 per 1,000 results ($0.005 per row) and a negligible per-run start fee.
• Every row counts as one result: 20 videos plus 200 comments is 220 results.
• Apify additionally bills the run's own compute time (one 2 GB browser). Measured, a 25-video run cost $0.0036 in compute — a fraction of a cent, not a hidden fee.
• Cap spend yourself with maxItemsPerKeyword or the run's maximum-cost setting. Comments are charged as ordinary rows, so decide deliberately before switching them on.

Integrate with your stack

• REST API — start a run and read the dataset with any HTTP client
• Python / JavaScript SDKs — pip install apify-client / npm i apify-client
• MCP server — point any AI agent at https://mcp.apify.com and let it call this Actor
• Webhooks and schedules — run it daily and diff on bvid to catch new uploads
• No-code — Google Sheets, Zapier, Make and the other Apify integrations

Why a real browser

Bilibili renders search results with JavaScript and rotates an internal API signature (wbi) that breaks lightweight scrapers within weeks. This Actor opens Bilibili's homepage so the site issues its own cookies, reads the rendered result cards, then calls Bilibili's view and reply endpoints through that same browser session. No reverse-engineered signature, no xsec_token chasing.

Bilibili scraper comparison

	This Actor	Typical Bilibili scrapers
Login required	Videos: never. Comments: one QR scan, reused on every later run	Sometimes, with no explanation
Search by keyword	Yes	Yes
Scrape exact video URLs	Yes	Rarely
Likes / coins / favorites / shares	Exact integers	Usually missing
Comments	Optional, priced per row	Text only, if at all
Danmaku (弹幕)	Yes, with timestamps and colour — no login	Almost never
Tags and description	Yes	Rarely
Play count	Integer (940040)	Rounded label (94.0万)
De-duplication	Automatic	Rarely
Price	$5 / 1,000 results — the same as the category leader	$5 – $20 / 1,000

How to log in to Bilibili (scan a QR - no DevTools)

A login is needed for uploader (userUrls) scraping, and for comments beyond 3 per video.

You never have to hunt for a cookie. The Actor draws a QR code and puts directions in three places, so you cannot miss it.

1. Start the run. The run's status line immediately changes to:

   SCAN TO LOG IN: open the Output tab and click 'Login QR code', then scan it with the Bilibili app

   That line is visible on the run page and in the Runs list — no clicking needed to read it.

2. Open the Output tab and click Login QR code. Three routes reach the same picture:

   Where	What you see	Good for
   The run's Output tab	An output called Login QR code	The route the status line tells you to take
   The run log	A ready-to-click link ending in records/LOGIN_QR	Fastest if you are already reading the log
   The run's Storage tab → Key-value store	A record named LOGIN_QR	If you closed the other two

3. Scan the picture with the Bilibili mobile app and tap confirm.

4. The run continues on its own — no restart needed.

The link changes every run, so always take it from the run you are watching.

You only scan once. After a successful scan the session is saved to a private key-value store named bilibili-session in your own Apify account, and every later run reuses it automatically — no scan, no cookie.

The Actor waits up to loginTimeoutSeconds for the scan. The default is 0 (never block); set it to 180 to be shown a QR. If nobody scans, it carries on anonymously, says so in the log and in SUMMARY, and — when failWithoutLogin is on — stops instead of returning partial data.

If the saved login ever stops working (you changed your password, logged out on the phone, or Bilibili expired it), the next run detects that, says so, and shows a fresh QR code. To force a re-login at any time, tick Forget the saved Bilibili login (resetStoredSession).

Why a QR and not a password? The Actor never sees, asks for, or stores your password. The session it receives is the same one your phone uses, and you can kill it any time by logging out in the app.

Advanced: paste a cookie yourself instead

If you prefer to run unattended with no scan, you can supply the cookie up front. Get it from your own browser:

1. Log in at bilibili.com in Chrome, press F12.
2. Open the Application tab — it is often hidden behind the » at the right end of the tab row.
3. Under Storage → Cookies, click https://www.bilibili.com.
4. Double-click the SESSDATA row's value cell, then Ctrl+C (⌘C on Mac).

┌─ DevTools ───────────────────────────────────────────────────────────────┐
│                                                                          │
│  Elements  Console  Sources  Network  »  [Application]                   │
│                                           ↑ click here                   │
│                                                                          │
│  ▼ Storage                                                               │
│      Cookies  ▸                                                          │
│          https://www.bilibili.com   ← click this                         │
│                                                                          │
│      name        value                     domain                        │
│      ───────────────────────────────────────────────────                 │
│      _uuid       8B2A…                     .bilibili.com                 │
│      buvid3      F1C2…                     .bilibili.com                 │
│      SESSDATA    7f3c9a4e2b1d8c05…         .bilibili.com  ← this one     │
│      bili_jct    9d41…                     .bilibili.com                 │
│      DedeUserID  2058285…                  .bilibili.com                 │
│                                                                          │
└──────────────────────────────────────────────────────────────────────────┘

5. Paste it into bilibiliCookie with the field name attached:

✅ SESSDATA=<paste-your-value-here>
❌ <paste-your-value-here>                (no field name)
❌ "SESSDATA=<paste-your-value-here>"     (quoted)

If this feels like too much, skip it. Use keywords instead — searching for the creator's name usually surfaces most of their videos, and it needs no login at all.

FAQ

Do I need a Bilibili account?

Depends on what you want:

You want	Account needed
Videos, search, uploader lists, plays, likes, coins, favorites, tags	No. Nothing to sign up for, nothing to renew
More than 3 comments per video	Yes — one free scan, remembered forever

Bilibili caps anonymous comment access at 3 per video. That is their rule, and no amount of retrying, proxying or endpoint-hopping changes it — only a logged-in session does.

What if my keyword has no matches?

Bilibili does not return an empty page — it falls back to fuzzy matches and recommended videos. Searching a nonsense term still produced five rows in testing, two of which happened to contain part of the query. That is Bilibili's own ranking, and those rows are billed like any other result. Filter on title or play_count in your spreadsheet if you need to drop them.

Can I scrape just a few specific videos? Yes — put the URLs or BVIDs in videoUrls and leave keywords empty. Stats are always fetched in that mode.

Why do stats differ from the search card? The search card shows a rounded cached label (94.0万); the stat block is the live counter from the video itself. Stats are accurate.

How many comments can I get?

Without logging in: 3 per video. That is a hard Bilibili limit, not a setting. Asking for 200 comments per video changes nothing while the run is anonymous — Bilibili answers HTTP 200 with a reply count of thousands and then hands back exactly three replies.

To get more:

• Log in. Set QR login wait (seconds) above 0 (for example 300) and scan the QR code with the Bilibili app. A logged-in session is what lifts the ceiling, up to maxCommentsPerVideo per video.
• Or paste a cookie into bilibiliCookie if you already have one.

Once logged in, Bilibili serves 20 comments per request (?ps=21 and above is rejected with ps out of bounds), and the Actor pages through them until it reaches maxCommentsPerVideo or hits Bilibili's rate limit. Per video the Actor reads at most 60 pages = 1,200 comments; top-level comments only, reply_count tells you how many replies each has.

Why the number looks smaller than the video's comment count. Bilibili's comment count adds top-level comments and their replies together. Measured on two videos with a logged-in session: 30 top-level + 50 replies = the 80 Bilibili displays, and 23 + 29 = the 52 displayed. The Actor returned all 30 and all 23 — i.e. every top-level comment, not a truncated sample. Use reply_count per row if you need to know how deep a thread is.

Two more things worth knowing:

• Bilibili meters its comment endpoints separately, and the paged endpoint starts answering HTTP 412 after a few hundred requests in a run.
• If both endpoints get rate-limited, comments stop for that run and SUMMARY.commentNote explains why. Video rows are never affected — comment scraping is deliberately independent of video scraping.
• When a run comes back with ≤3 comments per video, SUMMARY.commentNote tells you the anonymous cap is the reason, so you are never left guessing.
• If you skip the scan, the default behaviour is to carry on anonymously and return what it can. Set failWithoutLogin to true if you would rather the run stop with an error than hand you partial data.

If you do not have a Bilibili account, use this Actor for video and stats data (which needs no login at all) and treat comments as a bonus of 3 top comments per video.

Can I scrape all videos from an uploader (UP主)?

Yes — put their space URL (https://space.bilibili.com/123456) or bare numeric ID in userUrls. The run shows a QR code for you to scan with the Bilibili app; see How to log in above.

This mode needs that login. Anonymous requests to Bilibili's uploader endpoint get -799 (too frequent) or -352 (risk check), which is why a scan is required here and not for keyword search.

Large and official channels work too. Bilibili throttles the legacy uploader endpoint hard for big accounts — a 20-million-follower official channel answers -799 (too frequent) to every retry on it. This Actor therefore calls the signed endpoint the Bilibili web app itself uses (x/space/wbi/arc/search), and only falls back to the legacy one. Measured: that same official channel returned 50 videos in 74 s, then 80 videos across two pages in 39 s, all unique BVIDs.

If both endpoints ever refuse a channel, the run does not pretend it is empty: it logs an error, sets SUMMARY.uploadersFailed to the affected IDs, and explains why in SUMMARY.uploaderNote.

Why not just solve the captcha?

Because the captcha is not the real gate. Blocked requests come back as server-side risk control (-799 request-too-frequent, -352 risk check failed) — no image puzzle is ever shown for those. A logged-in session is what actually changes the outcome.

Can I scrape danmaku (弹幕)?

Yes, and unlike comments it needs no login at all. Turn on includeDanmaku and the Actor reads the video's danmaku XML directly, so you get the text, the exact second it appears (time_in_video), its colour, its mode and when it was sent. Set maxDanmakuPerVideo to control volume — a popular video can have thousands, and each danmaku is billed as one result row, so 200 is a sensible starting point.

Can I get replies to comments? Not currently — top-level comments only. reply_count tells you how many replies each has.

Is scraping Bilibili allowed? This Actor collects only publicly visible data. You are responsible for complying with Bilibili's terms of service and applicable law in your jurisdiction.

Something not working? Two ways to send us the run

Every run keeps a full log in the Apify console. Pick whichever is easier:

Option A - send the log file (best)

1. Open Actors → Bilibili Scraper → Runs and click the run that failed.
2. Open the Log tab (there is a download icon at its top right).
3. Save the .txt and send that file.

Option B - send the diagnostics link plus the log tail

1. Open the run's Output tab. The Diagnostics output is a signed link that opens without any token — paste that link as-is, or copy the whole DIAGNOSTICS JSON from the key-value store. It has the run ID, the input that was used, the row counts and the error flags.
2. Copy the last ~50 lines of the Log tab. The log itself is private to your account, so it cannot be linked — only copied or downloaded.

The signed links in the Output tab expose your scraped data to anyone holding the URL. Share them deliberately.

Either way we get everything needed. You do not have to send screenshots.

Note on privacy: the run log is private to your Apify account — the API URL for it needs your token. That is why the log has to be downloaded or copied rather than linked. DIAGNOSTICS never contains your bilibiliCookie.

What the usual symptoms mean

Symptom	What actually happened	What to do
0 videos	The keyword returned nothing, or everything was already seen	Try a different keyword; the log says "N new videos" per page
Videos fine, 3 comments per video	That is Bilibili's anonymous limit — not a bug	Set QR login wait above 0 and scan the QR to get more
Videos fine, 0 comments	Bilibili's comment quota ran out (HTTP 412)	Run comments in smaller batches, or accept the top-3 fallback
Uploader mode returns nothing	Nobody scanned the QR in time	Raise loginTimeoutSeconds, or paste a bilibiliCookie
Log says "stored session is no longer valid"	Bilibili expired your login	It shows a fresh QR — scan it again
Run fails immediately	Input problem	The log names the exact field

Keywords

bilibili scraper, b站 scraper, 哔哩哔哩, bilibili danmaku, 弹幕, bullet comments, 弹幕爬虫, bilibili search scraper, bilibili video scraper, bilibili comments scraper, bilibili api, bilibili data, bilibili up主, bilibili likes, bilibili stats, bilibili tags, bilibili dataset, chinese video platform data, bilibili analytics, bilibili export csv, bilibili json, bilibili up主 videos, bilibili uploader scraper, bilibili creator videos