Bilibili Scraper - Videos, Comments, Danmaku & Creators

Search Bilibili (哔哩哔哩, B站) by keyword, creator or trending board and get full video stats, top comments and raw danmaku as clean rows, with no login and at $2.00 per 1,000 videos.

Bilibili is where China's under-35 audience watches, reviews and argues about games, anime, phones, cars and brands, and it carries a signal no Western platform has: danmaku, the comments viewers fire across the screen at a specific second of the video. The data is public, but getting it out means signing every request the way Bilibili's web client does, dodging per-IP bans, and knowing which endpoints quietly hand an anonymous visitor three comments instead of three hundred. This Actor does that and gives you videos, comments, danmaku and creators as sortable rows, priced per row at a fraction of what the Store usually charges for Bilibili.

✅ What you get / ❌ what this isn't

✅ What you get	❌ What this isn't
Views, likes, coins, favourites, shares, comments and danmaku count on every video, plus an engagement rate	Not a title-and-link list
Every danmaku as its own row, with the second of the video it was sent at	Not a single summary row sold at $0.05 per video
A danmaku reaction profile on the video row: pool size, unique senders, peak minute	Not raw XML you have to parse yourself
Top comments ranked the way Bilibili ranks them, 20 per page, paged as deep as you ask	Not the 3 comments an anonymous visitor usually gets
Find creators (UP主) in a niche by keyword, sorted by followers	Not ID-only creator lookup
A creator's full upload list, newest first, paged to the end	Not "their videos that happen to show up in search"
Keyword search with sort, length and date filters that are checked to work	Not filters that are silently dropped
$2.00 per 1,000 videos, $0.50 per 1,000 comments, $0.10 per 1,000 danmaku, no start fee	Not $20 per 1,000 rows

🔎 Why use this Actor

• Priced per row type. Videos, comments, danmaku and creators each have their own price, set by what they cost to collect. A danmaku costs a tenth of a comment because one request returns hundreds of them, so a 200-danmaku sample of a video costs $0.02.
• Danmaku as data, not a teaser. Each danmaku row carries its text, videoSecond and m:ss timestamp, the time it was sent, a stable hashed sender id and its display mode. The video row adds danmakuPeakMinute, which tells you the minute of a review or ad the audience reacted to most. Set maxDanmakuPerVideo to 0 to get only that profile.
• Comments that go deep. Bilibili gives anonymous visitors a crippled comment feed of about three comments. This Actor requests the ranked feed the way that returns 20 per page and pages on until maxCommentsPerVideo, and flags pinned comments and comments written by the uploader.
• Creator discovery. creatorSearchQueries finds creators by niche keyword (美妆, 数码, 汽车) sorted by followers, with follower count, video count, level and verification, so you can build a KOL shortlist without knowing a single UID first.
• Complete upload lists. creatorIds returns a creator's profile and then pages through their uploads to the end, newest first, instead of guessing from search results.
• Filters verified against real results. All five sort orders, all four length buckets and the date window were each run and checked before release. A one-sided date window, which Bilibili ignores without saying so, is sent as a two-sided window so it actually applies.
• Survives Bilibili's IP bans. Bilibili bans an IP for a while once it has seen enough requests. The Actor spots the ban (HTTP 412 or risk-control codes) and moves to a fresh exit IP instead of returning an empty run.

👥 Who it's for

Brand and market researchers watching Chinese consumer reaction, influencer and KOL agencies, game and anime publishers tracking a launch, and data teams building Chinese-language datasets.

• Pull every review video of a product launch and rank them by engagement rate.
• See which minute of a competitor's ad or review the audience reacted to, from the danmaku timeline.
• Build a shortlist of beauty or tech creators by follower count, then pull their full upload history.
• Collect top comments on a game trailer to read what players actually objected to.
• Track the daily top-100 ranking and the live popular board on a schedule.

Example tasks

Ready-made runs you can open and start without configuring anything:

• Scrape Bilibili videos by keyword
• Scrape Bilibili comments on product reviews
• Scrape Bilibili danmaku with video timestamps
• Find Bilibili beauty creators (KOLs) by followers
• Find Bilibili gaming creators by followers
• Scrape all videos from a Bilibili creator
• Bilibili top 100 ranking (排行榜) today
• Bilibili popular videos with danmaku reaction profile
• Chinese gamer reaction to a game launch on Bilibili
• Monitor new Bilibili videos about a brand

⚙️ How to scrape Bilibili

1. Add one or more keywords to Search keywords. Chinese keywords return far more results than English ones (原神 rather than Genshin).
2. Or paste Video links, Creators (UID or space link), keywords to Find creators by keyword, or pick a Trending board. You can mix all of them in one run.
3. Set Maximum videos per keyword, creator or board. The default of 20 is a quick look; Bilibili serves up to about 1,000 per keyword.
4. Switch on Include comments and Include danmaku if you want them, and cap them per video.
5. Click Start. Rows stream in as each video finishes. Open the Output tab for the Videos, Comments, Danmaku and Creators tables, or export JSON, CSV or Excel from Storage.

Track a keyword or a board every day

Save your input as a task, set Sort videos by to Newest (or pick the Top 100 ranking board), and add a schedule. Each run returns the current state with fresh view, like and coin counts, so a daily export becomes a time series of how a launch or a creator is performing.

📥 Input

{
  "searchQueries": ["iPhone 17"],
  "maxVideosPerSource": 20,
  "sortBy": "totalrank",
  "fetchVideoDetails": true,
  "includeComments": true,
  "maxCommentsPerVideo": 40,
  "includeDanmaku": true,
  "maxDanmakuPerVideo": 200
}

• searchQueries - keywords to search videos for. Each keyword is searched separately.
• videoUrls - video links, BV ids, av ids or b23.tv short links.
• creatorIds - creator UIDs or space.bilibili.com links. Returns the profile plus uploads, newest first.
• creatorSearchQueries - keywords to find creators for, one row per creator.
• trending - none (default), popular (the live 综合热门 feed) or ranking (the top-100 排行榜).
• maxVideosPerSource - videos per keyword, creator or board. Default 20, up to 1000.
• maxCreatorsPerQuery - creators per creator-search keyword. Default 20.
• creatorSortBy - fans (default, most followers first), relevance or level.
• sortBy - keyword search order: totalrank (relevance, default), click (most viewed), pubdate (newest), dm (most danmaku), stow (most favourited).
• durationFilter - 0 any (default), 1 under 10 min, 2 10-30 min, 3 30-60 min, 4 over 60 min.
• publishedAfter / publishedBefore - date window for keyword search, YYYY-MM-DD, China time. Either side alone works.
• fetchVideoDetails - default true. Adds likes, coins, shares, originality, parts and engagement rate at one extra request per video. Turn off for a faster search-fields-only pull.
• includeComments / maxCommentsPerVideo - off by default; default cap 20 per video.
• includeDanmaku / maxDanmakuPerVideo - off by default; default cap 200 per video, sampled evenly across the timeline. 0 keeps only the reaction profile.

📤 Output

title	views	likes	coins	engagementRate	danmakuPeakMinute	creatorName
「小白」iPhone17全面测评：你想看的都在这了！	2,086,439	53,014	26,958	0.0479	-	小白测评
《原神》角色预告-「沃雅妮莎：此夜共沦」	2,099,652	259,693	101,311	0.2077	0	原神
和魔术师刘谦一起拉片魔术电影	7,346,132	320,422	128,642	0.0750	10	影视飓风

Every row has a type: video, comment, danmaku or creator, and the Output tab has a table for each. Video rows also carry bvid, aid, url, description, coverUrl, durationSeconds, publishedAt, category, tags, favorites, shares, comments, danmakuCount, isOriginal, parts, creatorId and creatorUrl, plus searchQuery, source and position so you know which input produced them. Comment rows carry the text, likes, replyCount, postedAt, the author's name, UID and level, and isPinned / isByUploader. Danmaku rows carry the text, videoSecond, videoTimestamp, sentAt, senderHash, displayMode and color. Creator rows carry followers, following, videoCount, totalLikes, level, isVerified, verificationTitle, bio and avatarUrl. A run summary with counts and anything that could not be fetched is saved as OUTPUT in the key-value store.

All output fields

Field	Row types	What it is
type	all	video, comment, danmaku or creator
source	video, creator	search, video-url, creator-uploads, popular, ranking, creator-id or creator-search
searchQuery	video, creator	Keyword that produced the row
position	video, creator	Rank within its keyword, creator or board
bvid	video, comment, danmaku	Video id (BV...)
aid	video	Numeric video id
url	video, creator	Page URL
title	video	Title
description	video	Description
coverUrl	video	Cover image
durationSeconds	video	Length in seconds
publishedAt	video	Publish time, ISO 8601 UTC
category	video	Bilibili zone
tags	video	Tags (keyword search rows)
views, likes, coins, favorites, shares	video	Engagement counts
comments, danmakuCount	video	Comment and danmaku totals on the video
engagementRate	video	(likes + coins + favourites + shares) / views
isOriginal	video	Original upload vs repost
parts	video	Number of parts
creatorId, creatorName, creatorUrl	video	Uploader
danmakuFetched, danmakuUniqueSenders	video	Danmaku pool size and distinct senders (danmaku on)
danmakuPeakMinute, danmakuPeakMinuteCount	video	Busiest minute and its danmaku count (danmaku on)
detailsFetched	video	Whether full stats were fetched
videoTitle, videoUrl	comment, danmaku	The video the row belongs to
commentId	comment	Comment id
text	comment, danmaku	Comment or danmaku text
likes, replyCount	comment	Comment likes and replies
postedAt	comment	Comment time
authorId, authorName, authorLevel	comment	Commenter UID, name and level (0-6)
isPinned, isByUploader	comment	Pinned by, or written by, the uploader
danmakuId	danmaku	Danmaku id
videoSecond, videoTimestamp	danmaku	Moment of the video it was sent at
sentAt	danmaku	When it was sent
senderHash	danmaku	Stable hashed sender id
displayMode, color	danmaku	scroll, top, bottom and so on; hex colour
creatorId, name, url, avatarUrl, bio	creator	Identity
followers, following, videoCount, totalLikes	creator	Audience and output
level, gender	creator	Account level and gender as shown
isVerified, verificationTitle	creator	Bilibili verification
isLive, liveRoomId	creator	Live status (creator search rows)
scrapedAt	all	Collection time

💵 How much does it cost?

You pay per row, by type, with no start fee, so a keyword that finds nothing costs nothing:

Row	Charge event	Price per 1,000 (Free plan)
Video	video-scraped	$2.00
Comment	comment-scraped	$0.50
Danmaku	danmaku-scraped	$0.10
Creator	creator-scraped	$2.00

Higher Apify plans get up to 20% off. A default run (20 videos for one keyword) costs $0.04 and finished in 13 seconds in testing. A heavy measured run - 100 videos each from a keyword, a creator's uploads and the top-100 ranking, with up to 40 comments and 200 danmaku per video - returned 296 videos, 11,108 comments, 48,965 danmaku and 21 creators in 22 minutes, which bills about $11. Comments and danmaku are off by default and each has its own per-video cap, so the bill only grows when you ask for it to.

🔁 Run it on the Apify platform

Save any input as a task and schedule it daily or weekly to watch a keyword, a creator or the ranking board. Call it from the Apify API or client libraries, export results as JSON, CSV, Excel or XML, and push them onward with webhooks or the Make, Zapier and n8n integrations, into a sheet, a warehouse or a Slack alert when a creator posts.

Use with AI agents via MCP

Add Apify's MCP server to Claude Code, Claude Desktop or Cursor, then ask the agent to run webdata_labs/bilibili-scraper with a keyword, a creator or a video link and to analyze the dataset. A research agent can chain it: search a product name in Chinese, pull the top comments and danmaku on the most-viewed reviews, then summarize what viewers praised or complained about at which minute. Keep your Apify token in the client's secret or environment settings rather than in a prompt.

Use this Actor in n8n

Use an HTTP Request node against the Actor's run endpoint (or run-sync-get-dataset-items for small runs), then map the dataset items into your workflow. Store the Apify token in n8n credentials, never inside the workflow JSON. A typical flow: a daily schedule trigger, a run over your brand keywords sorted by newest, a filter on views above a threshold, and a Slack or email message for each new video.

⚠️ Limits and caveats

• About 1,000 results per keyword. Bilibili serves at most 50 pages of 20 for one search. To go further, split the period with publishedAfter / publishedBefore windows or run related keywords.
• Comments are the top-ranked feed, not every comment. Anonymous access returns Bilibili's ranked comments at 20 per page. Replies under a comment are counted in replyCount but not returned as rows, and newest-first comment order is not offered because Bilibili gives anonymous callers only a handful that way.
• No commenter IP location. Bilibili shows a commenter's province only to logged-in users, so this Actor does not return it.
• Danmaku is the pool Bilibili keeps. A long, busy video may have sent more danmaku than Bilibili still serves; danmakuFetched is what the pool holds, danmakuCount is Bilibili's own lifetime count. On multi-part uploads the danmaku comes from the first part.
• Collaborations appear in a creator's upload list. Videos the creator co-uploaded with another account are included, with the lead uploader as creatorName.
• Upload-list rows are lighter unless details are on. A creator's upload list carries only title, views, duration and date; with fetchVideoDetails on (the default) each one is opened for full stats.
• Search rows without details lack coins and shares. With fetchVideoDetails off you get the search fields only: views, danmaku count, favourites, comments and tags.
• Region-locked, deleted or members-only videos return an entry in the run summary instead of a row.

🧩 Related Actors

• Weibo Scraper - the same brand or topic on China's microblog, to compare reaction across platforms.
• RedNote (Xiaohongshu) & Douyin Scraper - lifestyle and beauty trends on RedNote and Douyin, next to Bilibili's gaming and tech audience.
• 1688 Product Scraper - price the product a Bilibili review is about at its Chinese wholesale source.
• TikTok Shop Scraper - see whether the product Chinese viewers are reviewing is already selling on TikTok Shop.
• WeChat Official Account Scraper - long-form articles from brands and media on the same topic.
• TikTok Comments Scraper - the international short-video side of the same launch.

❓ FAQ

Do I need a Bilibili account, cookie or VPN?

No. Everything here is public and fetched without login. Proxy rotation and Bilibili's request signing are handled automatically and included in the price.

Am I charged when a keyword finds nothing?

No. There is no start fee and you are charged only per row delivered. Items that could not be fetched are listed in the run summary and never billed.

Why do my English keywords return so few videos?

Bilibili indexes Chinese text. Search 耐克 as well as Nike, 原神 as well as Genshin, and put both in searchQueries.

How do I get more than 1,000 videos for one keyword?

Run the keyword with several date windows (publishedAfter / publishedBefore), for example one per month. Each window has its own ~1,000 cap.

What is senderHash on danmaku?

Bilibili's own hashed sender id. The same viewer keeps the same hash, so you can count unique senders or spot one account flooding a video, but it does not reveal who they are.

Can I track a creator's followers over time?

Save a task with the creator in creatorIds and schedule it. Each run records the current follower, video and like counts with a timestamp.

I have input from another Bilibili Actor. Will it work?

Mostly, yes. searchQuery, keywords, queries, startUrls, urls, userIds, maxResults and maxComments are accepted as aliases for the fields above.

Is it legal to scrape Bilibili?

Scraping public data can be lawful in many situations, but legality depends on jurisdiction, purpose, platform terms and the data collected. This Actor collects only public video, comment, danmaku and creator information that Bilibili shows to any logged-out visitor, and never logs in or bypasses access controls. Comments and danmaku contain user-generated content and user names, so handle them under the privacy law that applies to you (for example GDPR or China's PIPL) and get legal advice for high-risk use.

Is this a Bilibili API?

Effectively yes. Run it through the Apify API, synchronously or asynchronously, and read structured JSON, CSV or Excel. The field names are stable and listed in the table above.

Changelog

• 2026-09-25 - First release: keyword video search with sort, length and date filters; video links; creator profiles with full upload lists; creator search by keyword; popular and top-100 ranking boards; full video stats with engagement rate; top comments; raw danmaku rows with a per-video reaction profile; per-row-type pricing with no start fee.

🛠️ Support

Something wrong or missing? Open an Actor issue with the run URL, the input you used and what you expected. That is the fastest route to a fix.

⭐ Rate this Actor

If this saved you time, please leave a rating on the Reviews tab. Review count is the main trust signal buyers have on the Store, and reviews decide which Actors get improved next. If something is broken, please open an issue first so it can be fixed.