🇨🇳 Bilibili Video Search by Keyword · B站视频搜索 (No Login)
Bilibili Video Search is a NexGenData tool that returns videos with BV-id, title, uploader, plays, likes and danmaku from Bilibili keyword search.
Turn any keyword into a ranked dataset of Bilibili (B站) videos — BV-id, title, uploader, play count, likes, danmaku (弹幕) count, duration, publish date and thumbnail — with no login, no official API key, and no cookie. Unlike URL-by-URL Bilibili scrapers, this actor is built around keyword search: feed it 人工智能 or "machine learning" and get the newest-and-hottest matches back as clean JSON or CSV. It serves two jobs especially well — assembling Chinese-language video/text training corpora for AI teams, and tracking what's trending with China's Gen-Z audience.
📊 Sample Output
Real rows from a real run (2026-09-25), unedited apart from masked emails/phones and shortened long text; fields the source does not publish are empty.
🔧 Input reference
| Field | Type | Default | What it does |
|---|
keyword (required) | string | "机器学习" | Keyword to search on Bilibili (Chinese or English). |
max_results | integer | 30 | Maximum videos. |
addonSentimentScored | boolean | false | OPT-IN. Score the (translated/transliterated) title + description with the VADER lexicon (positive / negative / neutral). Default OFF. |
addonSocialHandlesExtracted | boolean | false | OPT-IN. Fetch the creator's bilibili space page and regex-extract any linked Facebook / X / LinkedIn / Instagram / YouTube / TikTok / GitHub handles. Default OFF. |
🧾 JSON sample record
One real record from a real run (emails/phones masked, long text shortened):
{
"bvid": "BV1owrpYKEtP",
"aid": 113769053358965,
"url": "https://www.bilibili.com/video/BV1owrpYKEtP",
"title": "【官方中英】2025年公认最好的【吴恩达机器学习课程】附课件、代码及实战项目！！！--人工智能/机器学习/深度学习",
"description": "本课程字幕为官方英文字幕资源，由GPT翻译后经人工检查，最大程度上确保字幕适合食用，",
"author": "吴恩达AI_AndrewNg",
"mid": 3493134768016008,
"duration": "1384:56",
"play_count": 4423126,
"danmaku_count": 7757,
"favorites": 89761,
"likes": 33248,
"tag": "pytorch,人工智能,机器学习入门,深度学习入门,机器学习,学习,深度学习,吴恩达,卷积神经网络,计算机技术",
"pub_date": 1735979927,
"thumbnail": "//i0.hdslb.com/bfs/archive/3324ad7f3396ea7a8e363ae05b708ee3d3260436.jpg",
"keyword": "机器学习",
"scraped_at": "2026-09-25T00:03:28.492238Z"
}
中文说明
Bilibili 视频搜索让你用一个关键词批量获取 B 站视频数据,无需登录或 Cookie。每条结果包含标题、UP主、播放量、点赞、收藏、弹幕数与发布时间,可导出为 JSON、CSV 或 Excel,方便做品牌、市场与消费趋势研究。
定时运行:在 Bytewells 的「Schedules」中新建计划任务,设置关键词与数量上限,即可按天或按周自动追踪某一话题的热度变化。
说明:关键词搜索在匿名状态下可能受限;无结果时仅返回一条状态记录,不计费。所有字段均来自公开数据。
🤖 For AI agents & automation
An agent can research Chinese social and consumer trends on Bilibili — search a keyword and rank videos by engagement (plays, likes, danmaku) to gauge topic momentum. Note: a search that returns no results emits a single status record; keyword search may require a login cookie for some queries.
Sample agent prompt:
Search Bilibili for '智能手表' and return the top videos by play count with author and likes.
Output fields (one item per result):
bvid — Bilibili video idurl — video URLtitle — video titledescription — video descriptionauthor — uploader nameplay_count — view countdanmaku_count — on-screen comment countfavorites — favorite countlikes — like counttag — video tagspub_date — publish date
🧩 Optional add-ons (opt-in — you pay only if you switch them on)
Both are off by default. Switch either on in the input to enrich every result:
addonSentimentScored — adds a sentiment label + score (title + description) to each video, so you can rank results by audience mood with no post-processing.addonSocialHandlesExtracted — pulls creator/brand handles out of each description for outreach and lead lists.
Both default to off; leave them off and you pay only the base per-video price.
Use cases
- Build a Chinese-language video metadata corpus (titles + danmaku signals) for LLM / NLP training.
- Trend tracking — surging topics and creators on B站 for a keyword.
- Brand & campaign monitoring with China's Gen-Z audience.
- Equity / consumer research on Chinese media, gaming, anime, EV and streaming.
- KOL / creator discovery by topic, ranked by play count and danmaku.
Sample input
{ "keyword": "人工智能", "max_results": 100 }
keyword is required; max_results caps the number of videos returned.
Sample output
Each record returns bvid, title, author, play, like, danmaku, duration, pubdate, pic:
{ "bvid": "BV1xx411c7mD", "title": "10分钟看懂大语言模型", "author": "AI科技频道", "play": 482000, "like": 31200, "danmaku": 4870, "duration": "10:42", "pubdate": "2026-06-18" }
Run it on a schedule
For trend and brand tracking on Bilibili, a daily run per keyword is a good default, and weekly is enough for slower topics or corpus building. Between runs, new videos appear in the results and existing videos show updated play, like, favorite and danmaku counts, so each run is both new rows and fresh numbers. The actor has no new-only mode, so use bvid as the key to spot new videos and to track how engagement on known videos grows over time. Send each run's results onward with a webhook, Google Sheets, Zapier, Make or the Bytewells API.
FAQ
Why use this instead of Bilibili's own website or app?
Bilibili's site lets you search and watch, but it does not give you search results as data. This actor turns a keyword, in Chinese or English, into a list of videos as clean JSON or CSV, with BV-id, title, description, uploader, play count, likes, favorites, danmaku count, tags, duration, publish date and thumbnail. It needs no login, no API key and no cookie, it runs from outside China without a VPN on your side, and you pay per video.
What does a typical run cost?
One flat monthly price, with a free trial; there is no per-row charge. Bytewells bills compute separately from your prepaid credit, by memory and run time.
How do I search Bilibili videos by keyword and export the results?
Enter a keyword such as 人工智能 or machine learning and set max_results (default 30). Each video comes back as one row you can download as JSON, CSV or Excel, or pull through the Bytewells API. Keyword search is the core function, so you do not need video URLs first.
Does it return the danmaku comments themselves?
No. It returns the danmaku count for each video, which is a useful engagement signal, but not the comment text. For the comments under a video, use the Bilibili Comments Scraper.
Why did my search return no videos?
Bilibili can limit keyword search for visitors who are not logged in, and some queries may need a login cookie to return results. When a search finds nothing, the actor writes a single status record.
Can I use it to build a Chinese-language dataset for AI?
Yes. Titles, descriptions and tags across many videos per keyword give you structured Chinese-language metadata for a notebook, vector database or training pipeline. It returns metadata, not the video files or transcripts.
How it is billed on Bytewells
One flat monthly price for this actor, with a free trial to start. There is no charge per row, per result or per run from us.
Compute is billed separately by Bytewells from your prepaid credit, by memory and run time (per second). If a run uses the platform's proxy, Bytewells adds that traffic to the same run's compute bill.
Run it from the API
Bytewells speaks the same API as Apify, so curl or apify-client work with the Bytewells address and your Bytewells API key.
curl -X POST "https://bytewells.com/api/v2/acts/nexgendata~bilibili-video-search/runs" \
-H "Authorization: Bearer YOUR_BYTEWELLS_API_KEY" \
-H "Content-Type: application/json" \
-d '{"keyword": "机器学习", "max_results": 30}'
# when the run has finished, read its rows:
curl -H "Authorization: Bearer YOUR_BYTEWELLS_API_KEY" \
"https://bytewells.com/api/v2/actor-runs/RUN_ID/dataset/items"
from apify_client import ApifyClient
client = ApifyClient("YOUR_BYTEWELLS_API_KEY", api_url="https://bytewells.com/api")
run = client.actor("nexgendata/bilibili-video-search").call(run_input={'keyword': '机器学习', 'max_results': 30})
for row in client.dataset(run["defaultDatasetId"]).iterate_items():
print(row)
Contact
Questions, a source that changed, or a field you need: [email protected]