Bilibili Video Search, Ranking & Metadata Scraper

Bilibili Video Search, Ranking & Metadata Scraper

Collect public video metadata from Bilibili without an account. Search a keyword with Bilibili's own sort and filter options, resolve video links or IDs, walk the popular feed, or pull all-time, origin, rookie, and weekly ranking charts. Each row carries the identifiers, uploader, engagement counters, duration, publish time, cover, and optional detail fields such as tags, coin/share counts, parts, resolution, and uploader followers.

What this actor does

• Four modes: search, byVideoUrl, popular, and ranking (weekly charts are a ranking type).
• Search filters: sort by relevance, views, newest, danmaku, favorites, or comments; publication date range; duration bucket; and partition.
• Ranking coverage: 15 partitions across all-time, originals-only, rookie, and the weekly series (one issue per run, or a specific issue number).
• Detail enrichment: optional per-video enrichment adds tags, coin and share counts, cid, every part, resolution, uploader followers, and the full stat block. Video lookups are always enriched.
• Duplicate-safe: repeated videos across pages or queries are emitted once, and course-ad rows without a video ID are skipped.
• Stable field names: every mode uses the same names, and a field is omitted only when Bilibili has no real value for that video, never replaced with a placeholder.
• Empty fields are omitted: missing values are left out instead of being returned as null.

Output per video

• recordType — "video" for normal rows, "error" for error rows.
• bvid — BV-format video ID.
• aid — numeric AV video ID.
• cid — main part content ID (detail, popular, ranking, weekly, and enriched rows).
• title — video title, with search highlighting and HTML entities cleaned.
• description — video description.
• tags[] — video tags (search rows and enriched rows).
• typeName — partition or category label.
• partitionId — Bilibili category ID for the video (often a subcategory of the selected partition).
• author — uploader display name.
• authorMid — uploader numeric member ID.
• authorAvatar — uploader avatar URL.
• authorFollowers — uploader follower count (enriched rows).
• viewCount — total views.
• likeCount — likes.
• coinCount — coins (enriched, popular, ranking, and weekly rows; absent on raw search rows).
• favorites — favorites/collections.
• shareCount — shares (enriched, popular, ranking, and weekly rows; absent on raw search rows).
• danmaku — aggregate danmaku count.
• replyCount — comment count.
• duration — length in seconds.
• durationText — length formatted as H:MM:SS or M:SS.
• publishTime — publish time in UTC ISO format.
• publishDate — publish date in YYYY-MM-DD (UTC).
• coverUrl — cover image URL.
• pageCount — number of parts in a multi-part upload.
• pages[] — per-part objects with page, cid, title, duration, width, and height (enriched rows).
• resolution — main part resolution, for example 1920x1080.
• dynamic — the uploader's video dynamic text, when present.
• copyright — Bilibili copyright flag.
• uploadRegion — uploader's declared location, when published.
• rankIndex — 1-based chart position (ranking and weekly rows).
• recReason — recommendation reason shown by Bilibili (popular and weekly rows, when present).
• engagementRate — (likes + coins + favorites + shares + replies) / views, rounded to four decimals. Raw search rows have no coin/share counters, so their rate uses the counters Bilibili returns there.
• coinLikeRatio — coins / likes, rounded to four decimals, when likes are non-zero. Absent on raw search rows because Bilibili does not return coins in search results.
• videoUrl — canonical Bilibili video page.
• sourceUrl — the search page, chart, or video page the row came from.
• sourceQuery — original search keyword (search rows only).
• dataSource — search, video, popular, ranking, or weekly.
• enriched — true when per-video detail data was merged.
• scrapedAt — UTC timestamp of collection.

Output per error record

Error rows carry recordType: "error", an errorType (invalid_input, not_found, blocked, rate_limited, upstream_error, parse_error, or empty_response), a message, the sourceUrl that failed, optional statusCode and biliCode values, a retryable flag, and scrapedAt. Error rows never count toward maxItems.

Input

Either date bound can be used alone; the scraper fills in the missing bound because Bilibili applies a publish-date filter only when both bounds are supplied.

Example: newest short Python videos from the knowledge partition

{
  "mode": "search",
  "searchQueries": ["python"],
  "sort": "pubdate",
  "partition": "36",
  "duration": "short",
  "publishedAfter": "2026-01-01",
  "maxItems": 40
}

Example: enrich a batch of mixed video links

{
  "mode": "byVideoUrl",
  "videoUrls": [
    "https://www.bilibili.com/video/BV1wD4y1o7AS/",
    "BV1Jgf6YvE8e",
    "av170001",
    "https://b23.tv/abc123"
  ]
}

Example: top popular videos with full detail

{
  "mode": "popular",
  "maxItems": 100,
  "enrichDetails": true
}

Example: latest weekly chart in the food partition

{
  "mode": "ranking",
  "rankingType": "weekly",
  "partition": "211",
  "maxItems": 50
}

Use cases

• Creator and niche research — search Chinese and English keywords with date and duration filters to see who is active in a topic.
• Trend tracking — schedule daily pulls of a partition's all-time, origin, rookie, or weekly chart and compare rank movement.
• Content benchmarking — compare views, likes, coins, favorites, and comments across videos and partitions.
• Dataset building — export stable video metadata for recommendation models, topic analysis, or media monitoring.
• Catalog resolution — turn a list of saved links or BV IDs into complete metadata with one run.
• Engagement analysis — use the derived engagement rate and coin-to-like ratio to screen videos for authentic interaction.

FAQ

Do I need a Bilibili account or cookies?

No. Search, video lookups, the popular feed, and every ranking chart are read from public pages without a login.

Does this actor collect comments, danmaku text, subtitles, creator video lists, or video files?

No. The actor returns public video metadata and aggregate counters only. Comment text, individual danmaku messages, subtitles, creator video lists, and media downloads are outside its scope. Uploader details and follower counts come from the videos themselves.

Why are coinCount and shareCount missing on some search rows?

Bilibili's search results do not include coin or share counts. Set enrichDetails to true (or use byVideoUrl) to fill them from the video detail data.

How much slower is detail enrichment?

Each enriched video usually adds 3-10 seconds, depending on Bilibili latency. A 1,000-row search with enrichment can take an hour or more; the default run is intentionally small.

Why do I get fewer videos than maxItems?

Search returns at most 1,000 rows across 20 pages, the popular feed currently holds about 643 videos, a ranking chart has roughly 87-100 entries, and a weekly issue has about 48-49. Filters applied after collection can reduce the count further.

What do dm and scores mean in the sort list?

dm sorts by danmaku count and scores sorts by comment count. Older actors labelled scores as a rating; that is not what Bilibili returns.

Are course ads included in search results?

No. Search pages can contain course-promotion rows without a video ID; those are skipped, so a page of 50 results can yield fewer than 50 videos.

What timezone are the date filters in?

Both publishedAfter and publishedBefore are interpreted in UTC, and publishTime is emitted in UTC ISO format.

Can I fetch a specific weekly issue?

Yes. Set rankingType to weekly and pass the issue number in weekNumber. Leave weekNumber empty to get the latest issue.

What happens with an invalid link or ID?

That input produces an error row with recordType: "error" and an errorType, so you can see exactly which entry failed. Valid entries in the same run still produce video rows.

Do I need to configure a proxy?

No. The actor starts with a direct connection and retries through a fresh session and an anonymous browser session. If you provide a proxy, it is used only after those stages; without one, a persistent block ends in a typed error record.

Is this actor affiliated with Bilibili?

No. It is an independent third-party tool and is not affiliated with, endorsed by, or sponsored by Bilibili or its operators.