Node.js
The pdf-textlayer Node.js API — makeSearchablePdf, writeTextLayer, addTextLayer, loadPages, their TypeScript types, and error codes.
npm install pdf-textlayerpnpm add pdf-textlayeryarn add pdf-textlayerbun add pdf-textlayernub add pdf-textlayeraube add pdf-textlayerNode.js 20 or newer, ESM and CommonJS, with TypeScript types included. A prebuilt native module is installed for macOS (arm64, x64), Linux (x64 and arm64 with glibc, x64 with musl), and Windows (x64, arm64); nothing is compiled on install. Writes run on the libuv thread pool, so they never block the event loop.
import { makeSearchablePdf } from "pdf-textlayer";
const stats = await makeSearchablePdf("input.pdf", "parsed.json", "output.pdf", {
minConfidence: 0.5,
});const { makeSearchablePdf } = require("pdf-textlayer");
makeSearchablePdf("input.pdf", "parsed.json", "output.pdf").then(console.log);Functions
makeSearchablePdf
function makeSearchablePdf(
input: string,
source: Source | ParsedPage[],
output: string,
options?: MakeSearchableOptions,
): Promise<Stats>;
Reads OCR JSON with an adapter and writes the searchable PDF. source is a path (string or URL) to a JSON file, or the already-parsed document. It may also be an array of pages, in which case adapter is ignored.
writeTextLayer
function writeTextLayer(
input: string,
pages: ParsedPage[],
output: string,
options?: Options,
): Promise<Stats>;
Writes the PDF from boxes you already have. Pages not listed, or left with no drawable boxes, are copied unchanged.
addTextLayer
function addTextLayer(
pdf: Uint8Array,
pages: ParsedPage[],
options?: Options,
): Promise<{ pdf: Buffer; stats: Stats }>;
Same as writeTextLayer for a PDF held in memory, such as an upload or an object from storage.
import { readFile } from "node:fs/promises";
import { addTextLayer, loadPages } from "pdf-textlayer";
const pages = await loadPages("parsed.json");
const { pdf, stats } = await addTextLayer(await readFile("scan.pdf"), pages);
loadPages
function loadPages(source: Source, options?: { adapter?: Adapter }): Promise<ParsedPage[]>;
Resolves to the pages an adapter reads from OCR JSON, to inspect or edit before writing. Pages without OCR boxes are omitted, and malformed pages and items are skipped.
Types
interface TextBox {
text: string;
x: number;
y: number;
width: number;
height: number;
confidence?: number; // 0–1, default 1
rotation?: number; // degrees counter-clockwise, default 0
}
interface ParsedPage {
pageNumber: number; // 1-indexed page of the input PDF
boxes: TextBox[];
}
type FontName = "helv" | "tiro" | "cour";
type Adapter = "liteparse";
type Source = string | URL | Record<string, unknown>;
interface Options {
minConfidence?: number; // default 0
font?: FontName; // default "helv"
sanitizeUnsupported?: boolean; // default true
}
interface MakeSearchableOptions extends Options {
adapter?: Adapter; // default "liteparse"
}
interface Stats {
pages: number; // pages in the document
ocrPages: number[]; // 1-indexed pages that received a text layer
itemsDrawn: number; // boxes drawn across all pages
}
Boxes use PDF points with (x, y) the top-left corner on the displayed page; see Coordinates. The options match the Python options field for field.
Errors
Every function returns a promise that rejects with an Error whose code names what went wrong. Nothing is written when it rejects.
code |
Meaning |
|---|---|
ENOENT |
The input PDF or the JSON file does not exist. |
EIO |
Another file error, such as an unwritable output path. |
ERR_INVALID_PDF |
The input is not a readable PDF, or it is encrypted. |
ERR_INVALID_ARG_VALUE |
An unknown adapter or font, or JSON that does not parse. |
try {
await makeSearchablePdf("scan.pdf", "parsed.json", "out.pdf");
} catch (error) {
if (error.code === "ERR_INVALID_PDF") {
// skip files that are not PDFs, or are encrypted
}
}