RedNote image to text: get the text out of every image in a RedNote (Xiaohongshu) post
Short answer: on RedNote (小红书, Xiaohongshu) the content of most posts is written into the images: a one-line caption, then ten or fifteen picture cards of text. Copying the caption or fetching the page does not reach it. Paste the post's share link into LinkDigest and you get every image described and the text inside each one read out, in post order, as JSON or Markdown. No RedNote account, no cookies. 1 credit covers a post with up to 6 images, and each further 6 images add 1.
Disclosure: LinkDigest is the author's own service. The figures below come from a public demo page and a request made on 2026-11-02.
Why the text is out of reach
- It is pixels. Lists, prices, steps and quotes are typeset onto image cards. There is nothing to select or copy.
- The page does not hand it over. Requested without the app, the share link of the post used below was sent, on 2026-11-02, to RedNote's 404 page, whose only text is 「当前内容仅支持在小红书 APP 内查看」 — "this content can only be viewed in the RedNote app". Forty visible characters.
- AI chat cannot open the link. ChatGPT and most assistants cannot fetch a RedNote post at all, so they cannot read its images either.
One post, read
The public demo page shows what LinkDigest returned for one image post: a long article on self-improving AI, published as a run of text cards.
| Images read | 16, each described |
| Text found inside the images | 390 lines |
| Order | image 1, image 2, … as posted, each image's text kept with it |
| Price | 3 credits: 1 for the first 6 images, 1 for each further 6 |
The demo page publishes counts and one short sample only; the full text stays with the author's post. A read returns all of it to you.
What comes back
| Field | What it holds |
|---|---|
images[] |
One entry per image: description (what it shows) and ocr (the text inside it, verbatim) |
ocr_text[] |
Every distinct piece of text from all images, in one list |
caption, title, tags |
The post as written around the images |
key_points[] |
The takeaways, each with the quote it came from, checked against the post |
comments[] |
The top comments the share page shows |
coverage.images_read, images_total |
How many images were read, so a partial read is never silent |
Add translate_to: "en" and the image text comes back in English too, next to the Chinese — see translate a RedNote post.
How
In the browser: sign in and paste the share link on the workspace.
From code:
curl -X POST https://linkdigest.dev/api/v1/digest \
-H "Authorization: Bearer ld_live_..." \
-H "Content-Type: application/json" \
-d '{"url": "https://xhslink.com/o/...", "format": "markdown"}'
From an agent: the digest_url tool in Claude Code, Cursor or any MCP client; setup is on the Xiaohongshu platform page.
Copy the link in the app with Share → Copy link: RedNote serves a post only to a link that carries its xsec_token (why).
Limits, said plainly
- Up to 60 images are read from one post;
coveragesays if a post had more. - Handwriting and stylised fonts are read less reliably than printed text; each image's description says what it shows either way.
- Public posts only. A private or deleted post, or a link without its token, returns a clear error and costs nothing.
For a video post's speech, see RedNote transcript.