Bilibili Scraper: Video Search & Video Details (No Login)

Search Bilibili videos by keyword (up to 1,000 per keyword, five sort orders) or collect available public metadata for a video from its BV id, av id or URL. No login, no cookies: normal runs use plain HTTP, and a real browser is only the last fallback. 24 documented fields per video, null when the site does not provide a value. Result fees apply to saved videos. A separate memory-based Actor start fee also applies, including runs with no results. See Pricing below.

Sample item (video mode, description shortened for display):

{
  "bvid": "BV1c4411e77t",
  "aid": 52080698,
  "title": "【Python教程】《零基础入门学习Python》最新版（完结撒花🎉）",
  "description": "本系列教程是《零基础入门学习Python》最新版教程，希望大家喜欢。\n\n课后资源&作业 >>> BV1BJ41157PJ...",
  "url": "https://www.bilibili.com/video/BV1c4411e77t",
  "cover": "https://i2.hdslb.com/bfs/archive/383c952565090276d1908c1f50d416d0b43c8429.jpg",
  "durationSec": 63328,
  "publishedAt": "2019-05-10T22:33:01.000Z",
  "uploaderMid": 314076440,
  "uploaderName": "鱼C-小甲鱼",
  "uploaderUrl": "https://space.bilibili.com/314076440",
  "viewCount": 13846215,
  "likeCount": 283242,
  "coinCount": 227351,
  "favoriteCount": 683388,
  "shareCount": 119020,
  "commentCount": 10904,
  "danmakuCount": 144479,
  "tags": ["Python教学", "零基础Python入门", "Python3教程", "零基础入门学习Python第二版", "Python程序设计", "Python爬虫", "小甲鱼Python", "Python开发", "Python教程", "小甲鱼编程教学"],
  "categoryName": null,
  "position": 1,
  "sourceKey": "av52080698",
  "mode": "video",
  "scrapedAt": "2026-11-04T05:50:54.504Z"
}

Collection methods and limits

What	What you get
No account needed	Search and video pages are public. No login, no cookies, no signature handling on your side.
HTTP first	Measured: about 21-25 KB of proxy traffic per search page (about 37 videos) and about 90 KB per video document. A browser is started only if both HTTP paths fail.
Three acquisition paths	Search API, then the server-rendered search page, then a browser. The path used for every page is recorded in the RUN_SUMMARY record. The API gives the most complete and stable lists; fallbacks may recover some blocked requests, but access is not guaranteed.
Exact, documented schema	24 keys, exact types, the same keys in both modes, null (or []) when a value is unavailable.
Null is not success	A video without id, title, uploader id, play count or publish date is never written to the dataset and never charged.
Chinese as served	Titles, descriptions and tags come back in the original Chinese (zh-CN), not machine-translated.
Honest window information	Bilibili exposes at most about 1,000 results per keyword and pages overlap. RUN_SUMMARY reports numResults, numPages, the real count and windowLimited.
Drift detection	RUN_SUMMARY carries a hash of the raw key structure, so a site layout change shows up as a changed hash.

Modes

• search: keywords (1-50), order and maxItems (1-1,000 per keyword). Each keyword is searched on its own; duplicates inside a keyword are removed; position is the rank among the videos you receive.
• video: videoUrls (1-1,000): BV ids (BV1c4411e77t), av ids (av52080698) and bilibili.com/video/... URLs in any mix. The same video given twice (for example as BV id and as av id) is returned and charged once. Videos that are deleted, invisible or refused get a documented reason in RUN_SUMMARY.failed (for example video-not-found) and are not charged.

Sample item (search mode; coinCount and shareCount are only available in video mode, description shortened for display):

{
  "bvid": "BV1vGxF6sEsz",
  "aid": 117523710218309,
  "title": "【全500集】 目前2026最全最细的Python零基础全套教程，（包含爬虫+数据分析+AI）一周逼自己学完！从入门到精通，看完这一套就够了！",
  "description": "本套教程为2026最新录制，非常适合毫无编程经验的零基础小白，用来学习提升自己...",
  "url": "https://www.bilibili.com/video/BV1vGxF6sEsz",
  "cover": "https://i2.hdslb.com/bfs/archive/40e04b3f9a0eff07d50f6a40b4fa08baad80522a.jpg",
  "durationSec": 370415,
  "publishedAt": "2026-10-29T10:31:14.000Z",
  "uploaderMid": 3546801739139971,
  "uploaderName": "Python入门零基础教程君",
  "uploaderUrl": "https://space.bilibili.com/3546801739139971",
  "viewCount": 13644,
  "likeCount": 615,
  "coinCount": null,
  "favoriteCount": 1541,
  "shareCount": null,
  "commentCount": 127,
  "danmakuCount": 327,
  "tags": ["人工智能", "编程", "计算机", "Python安装", "Python教程", "Python基础", "Python爬虫", "Python入门", "Python", "数据分析"],
  "categoryName": "计算机技术",
  "position": 3,
  "sourceKey": "python 教程",
  "mode": "search",
  "scrapedAt": "2026-11-04T05:48:29.546Z"
}

Input

Field	Type	Default	Notes
mode	search or video	search
keywords	string[]	[]	Required in search mode, 1-50 entries. Chinese keywords work best.
order	relevance, pubdate, click, dm, stow	relevance	Newest first, most played, most danmaku, most favorited. Checked on 2026-11-04 for one keyword with 42-50 results each: no order violation for any of the four sorted orders.
maxItems	integer 1-1,000	100	Per keyword.
videoUrls	string[]	[]	Required in video mode, 1-1,000 entries.
maxConcurrency	integer 1-10	3	Parallel keywords or videos. Requests are also spaced out.
proxyConfiguration	proxy	Apify residential, US exit	Datacenter proxies were refused by video pages (HTTP 412) in the test, so video mode needs residential IPs.
debug	boolean	false	Verbose log; bodies of unparseable responses are saved as DEBUG_* records.

Invalid input fails immediately with an Input: message (for example Input: maxItems must be an integer between 1 and 1000 (got 2000).) and no video-result fee is charged. The Actor start fee may still apply.

Input examples

1. 100 relevant videos for one keyword

{ "mode": "search", "keywords": ["python 教程"], "maxItems": 100 }

2. Newest videos for several keywords

{ "mode": "search", "keywords": ["耳机 评测", "机械键盘"], "order": "pubdate", "maxItems": 50 }

3. As many videos as the site exposes, most played first

{ "mode": "search", "keywords": ["原神"], "order": "click", "maxItems": 1000 }

4. Full metadata (including coin and share counts and all tags) for known videos

{ "mode": "video", "videoUrls": ["BV1c4411e77t", "av52080698", "https://www.bilibili.com/video/BV1xx411c7mD/"] }

5. Explicit proxy and concurrency

{ "mode": "search", "keywords": ["教程"], "maxItems": 200, "maxConcurrency": 5,
  "proxyConfiguration": { "useApifyProxy": true, "apifyProxyGroups": ["RESIDENTIAL"], "apifyProxyCountry": "US" } }

Output fields

Every item has the following 24 keys. The table below describes their meaning and availability.

Field	Type	Example	Nullable	Notes
bvid	string	BV1c4411e77t	no	Video id.
aid	integer	52080698	yes	Legacy numeric id.
title	string	【Python教程】...	no	Highlight markup removed, HTML entities decoded.
description	string	本系列教程是...	yes	As served; can be an empty string.
url	string	https://www.bilibili.com/video/BV1c4411e77t	no	Canonical URL.
cover	string	https://i2.hdslb.com/bfs/archive/....jpg	yes	Always https.
durationSec	integer	63328	yes	Seconds.
publishedAt	string	2019-05-10T22:33:01.000Z	no	ISO-8601 UTC.
uploaderMid	integer	314076440	no	Uploader member id.
uploaderName	string	鱼C-小甲鱼	yes
uploaderUrl	string	https://space.bilibili.com/314076440	yes
viewCount	integer	13846215	no	Plays.
likeCount	integer	283242	yes
coinCount	integer	227351	yes	Video mode only; null in search mode.
favoriteCount	integer	683388	yes
shareCount	integer	119020	yes	Video mode only; null in search mode.
commentCount	integer	10904	yes	Number of comments (the comments themselves are not collected).
danmakuCount	integer	144479	yes	Bullet comments.
tags	string[]	["Python教学"]	no ([])	Complete in video mode, partial in search mode.
categoryName	string	计算机技术	yes	Search mode only; null in video mode (the video page serves an empty category name).
position	integer	3	no	Rank within the sourceKey; always 1 in video mode.
sourceKey	string	python 教程	no	The keyword, or the BV id / av id / URL as you gave it.
mode	string	search	no	search or video.
scrapedAt	string	2026-11-04T05:38:09.022Z	no	ISO-8601 UTC.

RUN_SUMMARY (key-value store)

Written at the end of every run: counts (itemsPushed, itemsRejected, itemsDuplicate, blockedEvents, retries), pages per path (pathPages: api, ssr, browser, http), a record for every page (pages: key, page, path, rows, new items, attempts, milliseconds), per keyword or video (perKey: status, numResults, numPages, pagesize, firstPageCode, windowLimited, endReason, rawKeyHash), nullRateByField, rawKeyHash (listing and detail), failed (key, url, reason, retries) and chargeLimitReached.

Use cases

• Track what is trending for a topic: search with order: pubdate or click on a schedule and compare viewCount between runs.
• Build a dataset of Chinese video titles, descriptions and tags for research or model training (public metadata only).
• Look up a list of known videos: statistics (plays, likes, coins, favorites, shares, comments, danmaku) and tags in one pass.
• Creator and content research: find the uploaders (uploaderMid) behind the videos that rank for your keywords.

Pricing

Event	Price	When it applies
video-result	$0.004 per saved row ($4 per 1,000)	Keyword-search videos saved to the dataset.
video-detail	$0.008 per saved row ($8 per 1,000)	Video-detail records saved to the dataset.
apify-actor-start	$0.00005 per start event	Apify counts one event per GB of allocated memory, with a minimum of one event.

At the current default 4 GB, the start fee is $0.00020; 100 search results therefore cost $0.40020 in Actor events. Use the Pricing tab for current prices and check the memory allocation before running. There is no additional automatic dataset-item fee. Platform compute and Apify proxy usage during the run are included; later storage, API use or integrations follow your account plan.

The start fee can apply even when input is invalid, requests are blocked or no results are returned. Dropped videos have no result fee. Rows already saved remain billable if the run later fails; inspect RUN_SUMMARY and the dataset before retrying. Set a maximum charge limit per run. Budget checks and writes are serialized across workers. A run with existing output cannot be resumed or resurrected; start a new run and treat it as a separate, billable collection.

FAQ

Will it get blocked? Sometimes the search API refuses a request even with the right headers. In the verification runs of 2026-11-04 the API answered normally until about 15:30 KST (after roughly 55 API requests from the test machine in the previous hour, including a 31-page burst), then answered code 0 with no results and only a risk-control voucher, first for some and then for every request from that machine's IP, and for 5 of 7 requests through a US residential exit. The Actor treats that "soft block", HTTP 412 and the JSON codes -352 / -403 as a block: it retries on a fresh proxy session (a new exit IP), then switches to the server-rendered search page, then to a browser. The runs during the soft-block period still completed from the server-rendered page. Every retry and switch is logged and counted in RUN_SUMMARY (blockedEvents, retries, pathPages). Video documents: 30 requests through the US residential exit, one HTTP 412, which succeeded on the automatic retry. A run never crashes because of one keyword; if no item at all is produced it ends with an error and a clear message.

How fast is it? Measured on 2026-11-04 through a US residential exit with the default concurrency of 3: 100 search results in 12.6 s (and 35.1 s in a later run: the speed of the residential exit varies a lot), 977 results for one keyword in 63 s, 12 video documents in 32 s (22.7 s in a later run). From a home connection without a proxy (KR) the same 100 results took 1.8 s.

How many items per keyword? Up to 1,000. Bilibili exposes at most about 1,000 results per keyword (24 pages of 42 rows) and pages overlap, so a broad keyword returns a bit fewer unique videos: 977 for 原神 through the API. If the API is refusing and the Actor falls back to the server-rendered page, that page is re-ranked on every load and overlaps more: the one 24-page test returned 662 unique videos for 原神 (346 duplicates removed). RUN_SUMMARY.perKey[].windowLimited, itemsDuplicate and pathPages show which case you got.

Which countries and proxies? The default is Apify residential with a US exit and all verification used it. Datacenter proxies served search but were refused by video pages with HTTP 412 in the test. Other residential countries are untested (experimental).

Is the text in English? No. Bilibili serves Chinese; the Actor asks for the Chinese version explicitly so titles, descriptions and tags are not machine-translated.

Are comments, uploader pages or danmaku included? No. Anonymous access to comments is capped by the site, so comments are not part of this Actor. Uploader-page collection is not supported.

What happens to deleted or invisible videos? They appear in RUN_SUMMARY.failed with a reason such as video-not-found and are not charged.

Legal note

The Actor reads public web pages and public API responses without logging in. It collects video metadata only: no comments, no commenter data, no private account data. Public uploader names, IDs and URLs are included and can identify individuals. Uploader name, id and URL are part of the public video record. You are responsible for using the data lawfully (copyright, platform terms, privacy and data-protection law in your jurisdiction).

Changelog

2026-11-06: clarified start fees and partial-run billing, serialized budget checks and writes, and blocked restarts that already contain output. Historical measurements above describe specific test runs; they are not speed or availability guarantees.