Social URL to LLM Text

Paste a Xiaohongshu, Douyin, YouTube or X link and get back text a language model can actually read: transcript, on-screen text, image descriptions, caption and metadata.

Official website: linkdigest.dev — powered by LinkDigest, which also offers this as an MCP server and a REST API.

The problem this solves

Fetching a social link yourself returns nothing useful. Share URLs are tokenised (xsec_token, app_code_link), the content lives in video and images rather than HTML, and the server sends an app-download shell or a login wall instead of the post. There is nothing in the page to parse.

$ curl -s 'https://xhslink.com/o/1WiQ1QI6Uc0' | grep -o '<title>.*</title>'
<title>小红书</title>

That is the whole page. No caption, no images, no text.

This Actor returns the post instead:

{
  "platform": "douyin",
  "author": "逸蒙的赛博空间",
  "title": "这只丑萌AI竟然月售100万美金！",
  "posted_at": "2026-08-27",
  "key_points": [
    "该AI产品名为 Tolen，主打「记住你」的核心功能",
    "商业模式为订阅制，提供免费体验，付费解锁更长时间的聊天服务"
  ],
  "transcript": [{ "t": 0, "text": "谁能想到，就这么一个丑萌小玩意儿…" }],
  "transcript_source": "asr",
  "images": [{ "description": "…", "ocr": "…" }],
  "ocr_text": ["…"],
  "credits": 7
}

Platforms

Every row below was checked against a real post, not a documentation page.

Platform	Status
Xiaohongshu 小红书	Works — including image notes, with every image described and OCR'd
Douyin 抖音	Works — video transcribed
YouTube	Works — native captions where available, ASR otherwise
X	Works
Web pages / articles	Works
TikTok	Not supported. TikTok blocks our servers' IP range
Bilibili	Not supported. Returns HTTP 412 to our servers
Instagram, Facebook	Not supported

The two unsupported rows are listed on purpose. Finding out after you have wired something in is worse than knowing now.

Input

Field	Type	Notes
urls	array of strings	Required. Share links work as-is.
format	json or markdown	json gives separate fields; markdown gives one ready-to-paste document per link.
maxItems	integer	Optional ceiling so a long list cannot cost more than expected.

Output

One dataset item per link. In json mode:

platform, author, title, posted_at, caption, transcript ({t, text}), ocr_text, images ({description, ocr}), key_points, source_url, transcript_source (native_captions | asr | none), degraded, cached, credits.

A link that cannot be read still produces an item, carrying url and error. Nothing is silently dropped — a short dataset is the worst way to discover that three of your fifty links failed.

Pricing

Event	Price
digest-completed	$0.02 per link successfully read
extra-credit	$0.01 per credit beyond the first

Credits track the two things that actually cost money: images to describe and minutes of media to transcribe. One base credit covers a typical post and its first six images; each further six images adds one, and each minute of audio or video adds two.

Example	Price
A web article or short text post	$0.02
A Xiaohongshu note with 17 images	$0.04
A 2.5-minute Douyin video	$0.08
A 19-minute YouTube video	$0.40

A link that fails is never charged. Neither is a run that fails because of a problem on our side.

How long it takes

Real measurements, not estimates:

• Already-digested link: about 1 second (anything anyone has run before is cached)
• Xiaohongshu note with images: 1–2 minutes
• YouTube video with captions: about 2.5 minutes

Links are processed four at a time.

Notes

• The heavy work runs on our servers, not on Apify. That is deliberate: the platforms above refuse datacenter IP ranges, so fetching from inside a cloud Actor is exactly what does not work.
• Chinese text is returned as-is, not translated.

Also available at linkdigest.dev

The same service runs at linkdigest.dev, where you can use it three other ways:

• MCP server — inside Claude Code or Cursor, so an agent can read a link mid-conversation:

  claude mcp add --transport http linkdigest \
    https://linkdigest.dev/mcp \
    --header "Authorization: Bearer ld_live_..."
  

• REST API — POST https://linkdigest.dev/api/v1/digest
• Web app — paste a link and read the result in the browser

Three digests free, no card. Questions or a link that should work but doesn't: linkdigest.dev.