Accessibility
Example
// Only when the host provides its own accessible representation of the document. new PDFViewer(container, doc, { accessibility: { textLayer: false } });
Options 1
Accessible text layer over each page. Screen readers read pages in reading order (each page is an ARIA region labeled "Page N of M"): the text the viewer is designed to expose for a WCAG 2.1, Section 508 or EN 301 549 review. Native find-in-page matches text on rendered pages; Ctrl/Cmd+F opens the SDK's own Search, which covers the whole document, and the browser's find gets the key when Search is off (tools.search: false) or shortcuts are (navigation.keyboardShortcuts: false). Native selection & copy work with no tool active (defaultTool: null) — drag-select and Ctrl/Cmd+C like regular web text; the selection highlight aligns with the rendered glyphs. Select, the resting tool by default, selects and copies the same text with its own selection.
Set it to false only when the host application provides its own accessible representation of the document.
License. The text layer puts the page's words into the page's markup, where any script on the page can read them, so it is built only under a license that includes extract or whose output is clean (a production license). Under a development key without extract, an unverified or an unlicensed session, the pages are shown without it.
Notes
The layer is built lazily for pages near the viewport and rebuilt automatically after every document change (edits, undo/redo, rotation, page management). While any tool is active the layer is pointer-inert — tools own all pointer interaction — but its text stays in the accessibility tree. Pages whose "text" is vector outlines (no text objects) have nothing to expose; that content is not readable by any PDF text extractor.
Tagged documents. A tagged PDF (the kind PDF/UA, the accessibility standard, requires) says what each drawing on a page is, so a screen reader reads the document in order. What the SDK draws keeps it so. Text added with the Add text tool becomes a paragraph of the document's structure, and a picture or a Sign-tool signature a figure, read after what is already on its page; a signature is read out as "Signature" (the signature.alt string), while a picture gets no description the SDK cannot know — give it one in an accessibility editor. Edited text stays where it was in the structure. What is not the document's own words — the license watermark, a black-out box — is marked as an artifact, which screen readers pass over. An untagged document is left untagged.
Keyboard access to the toolbar, panels and dialogs is described under Keyboard shortcuts.