Bilibili Transcript Scraper — AI Speech-to-Text for B站 Videos

Turn any Bilibili (B站) video into text with real AI speech recognition. Paste video links, get back the full spoken transcript plus timestamped sentences as clean JSON — with best-in-class Mandarin Chinese accuracy, ready for LLMs, RAG pipelines, content research, and subtitle workflows.

This is real ASR, not subtitle scraping. Most "transcript" tools only download subtitles a creator happened to upload — and the majority of Bilibili videos have none. This Actor runs industrial speech recognition on the video's audio track, so it works on any Bilibili video with speech, subtitles or not. Chinese (Mandarin) recognition is its strongest language.

Unofficial. This Actor is not affiliated with, authorized, or endorsed by Bilibili. It is an independent tool that processes publicly available content. Use it in compliance with Bilibili's terms and all applicable laws; you are responsible for how you use the retrieved data.

What you get per video

• Full transcript (fullText) — the complete spoken content as one string.
• Timestamped sentences (sentences[]) — each with startMs / endMs, ready for subtitles, deep links, and clip selection.
• Video metadata — title, UP主 (author), duration, publish time, view / like / coin / favorite / danmaku / comment / share counts, and cover image URL, so every transcript arrives with its engagement context attached.

Quick start

1. Open the Actor and press Run — the default input works out of the box.
2. Replace the example with your own video links: full URLs (https://www.bilibili.com/video/BV…), short share links (https://b23.tv/XXXX), or bare BV ids.
3. Language defaults to Chinese; switch to Auto for mixed or non-Chinese content.
4. Collect results from the Dataset tab as JSON / CSV / Excel, or pull them via the Apify API and MCP from your own code.

No proxies, no cookies, no login — everything runs server-side.

Example output

{
  "videoId": "BV1HjmrBzEDA",
  "videoUrl": "https://www.bilibili.com/video/BV1HjmrBzEDA/",
  "title": "Ultra级旗舰!OPPO Find X9 Pro好用到超标 #数码生活 #科技数码",
  "author": "科技小辛",
  "durationSec": 87,
  "viewCount": 152000,
  "likeCount": 8400,
  "coinCount": 1200,
  "danmakuCount": 430,
  "language": "zh",
  "detectedSpeech": true,
  "sentenceCount": 13,
  "fullText": "如果有一台手机能让果粉变墨粉，那一定是OPPO Find X9 Pro…",
  "sentences": [
    { "text": "如果有一台手机能让果粉变墨粉，", "startMs": 280, "endMs": 2350 },
    { "text": "那一定是OPPO Find X9 Pro。", "startMs": 2350, "endMs": 4400 }
  ]
}

What people build with it

• Chinese-market research — index what B站 UP主 actually say (not just titles) to understand trends, opinions, and product sentiment in your category.
• Long-form knowledge mining — Bilibili is China's home of tech reviews, tutorials, and video essays; transcripts turn hours of video into searchable, quotable text.
• Viral-hook mining — pull transcripts of the top videos in your niche and study the exact opening lines and structures that earn views and coins.
• Cross-border content — transcribe Chinese videos, then translate and repurpose them for YouTube or your own market.
• LLM & RAG pipelines — build Chinese-language corpora from real video speech, with engagement scores as a free quality signal.
• Subtitles & translation — timestamped sentences drop straight into SRT/VTT generation and dubbing workflows.

Pricing & billing

Pay per started transcript minute ($0.025/minute, one-minute minimum). Failed fetches, deleted or region-locked videos, videos over your duration cap, and failed transcriptions are not charged.

Why this Actor

• Best-in-class Chinese ASR — Mandarin recognition is its strongest language, where Western transcript tools struggle most.
• Works without subtitles — real speech recognition on the audio track; most Bilibili videos carry no CC subtitles to scrape.
• Audio-only downloads — Bilibili serves audio as a separate stream, so the Actor fetches megabytes instead of the full video. Long videos transcribe fast and cheap.
• Engagement context included — every transcript ships with view / like / coin / favorite / danmaku counts, so you can rank by performance immediately.
• Direct API, no headless browser — fast, stable runs with nothing to babysit.
• No login, no cookies — we never touch your accounts, so there's no ban risk.
• Structured JSON — export to CSV, Excel, or JSON, or pull straight from the API / MCP.

Tips for better results

• Feed it the winners: find the top-performing videos in your niche first (search / channel scrapers), then transcribe just those.
• Keep the language on Chinese for B站 content — it noticeably improves accuracy over Auto.
• The default duration cap is 60 minutes; raise maxDurationMinutes (up to 240) for lectures and documentaries, or lower it to keep batch runs snappy.
• For multi-part videos (分P), part 1 is transcribed; submit other parts' direct URLs if you need them.
• detectedSpeech: false flags music-only / no-speech videos so you can filter them out downstream.

FAQ

Do I need an account, cookies, or to log in anywhere? No. The Actor talks to fast, direct HTTP APIs server-side — you just provide video links and run it.

Does it work on videos without subtitles? Yes — that's the point. It runs real speech recognition on the audio, so subtitles are never required (and most Bilibili videos don't have them).

How good is the Chinese accuracy? Mandarin is the model's strongest language; it handles fast colloquial speech, regional accents, and technical vocabulary well, and returns punctuated sentences.

What about really long videos? Videos longer than your maxDurationMinutes cap (default 60) are skipped and not charged. Raise the cap up to 240 minutes when you need lectures or documentaries — audio-only downloads keep even those runs quick.

How am I billed? $0.025 per started transcript minute. Videos that can't be fetched or transcribed are not charged.

Can I run it on a schedule or call it from my app? Yes — use Apify Schedules, the REST API, the JavaScript / Python clients, or the MCP server. See the API tab.

Is this affiliated with Bilibili? No. It's an independent tool that processes publicly available content. Use it in line with the platform's terms and applicable law.