---
title: "Text extraction"
description: "Extract a page's plain text with the per-page text service."
source: "https://www.cloudpdf.com/docs/engine/core-concepts/text"
---

# Text extraction

Each page handle exposes a `text` service that returns the page's full plain-text
extraction.

```ts
interface PageTextService {
  read(): AbortablePromise<PageTextSnapshot>;
}
```

## Reading text

```ts
const page = doc.page(pageObjectNumber);
const { text, charCount } = await page.text.read();

console.log(text);      // full page text in display order
console.log(charCount); // PDFium-reported character count
```

The result is a `PageTextSnapshot`:

```ts
interface PageTextSnapshot {
  text: string;
  charCount: number;
}
```

- `text` is the page's text in display order, decoded from UTF-16 to a JS
  string.
- `charCount` is PDFium's character count. It can differ from `text.length`
  when the page contains astral-plane characters: PDFium counts UTF-16 code
  units, and JS strings keep them as surrogate pairs.

> The text snapshot is pure content — it's addressed and cached by the page's
> content version on the server, so repeat reads of an unchanged page are fast
> and CDN-friendly. It deliberately carries no annotation liveness; that lives
> on [annotation reads](https://www.cloudpdf.com/docs/engine/core-concepts/annotations).

## Extracting a whole document

There's no document-level text call — iterate the page list and read each page:

```ts
const { pages } = await doc.pages.list();

const fullText = (
  await Promise.all(pages.map((p) => doc.page(p.pageObjectNumber).text.read()))
)
  .map((snapshot) => snapshot.text)
  .join('\n\n');
```

> Reading text requires the `doc.text.copy` capability on the cloud.
> If the caller's scope doesn't grant it, `read()` rejects with a
> `Forbidden` `EngineError`. Check
> `doc.security.effectiveScope` before exposing a "copy text" action.