pdf-raster

Blazing fast, native PDF-to-image conversion for Node.js and Bun.

pdf-raster renders PDF pages into high-quality image buffers using a high-performance Rust engine. Built with napi-rs and PDFium, it is optimized for server-side workflows like OCR pipelines, VLM (Vision Language Model) inputs, and document previews.

Explore Documentation

---

✨ Why pdf-raster?

• 🚀 High Performance: 4x–8x faster than pdfjs-dist canvas-based implementations in the included local benchmark run.
• 🦀 Native Power: Leverages Rust and PDFium for industry-standard rendering.
• 🛠️ Simple API: A single convert() function handles everything.
• 📦 Memory Efficient: Supports Buffer inputs for processing without temp files.
• 🎨 Flexible Output: Native support for png, jpeg, and webp.

📦 Install

# Using Bun
bun add pdf-raster

# Using PNPM
pnpm add pdf-raster

# Using NPM
npm install pdf-raster

📖 Quick Usage

import { convert } from "pdf-raster";
import { writeFile } from "node:fs/promises";

// Convert specific pages to high-quality WebP
const [page] = await convert("./report.pdf", {
  pages: [0], // 0-indexed page numbers
  dpi: 300, // Higher DPI is useful for OCR and VLM inputs
  outputFormat: "webp",
});

console.log(`Rendered page ${page.pageIndex} (${page.width}x${page.height})`);

// page.data is a Buffer of the encoded image
await writeFile("output.webp", page.data);

🛠 API Reference

convert(input, options?)

input

• Type: string | Buffer | Uint8Array | ArrayBuffer
• Description: A file path to a PDF or an in-memory PDF byte source.

options (Optional)

Option	Type	Default	Description
pages	number[]	all	Specific pages to render, using 0-indexed values.
dpi	number	300	Resolution of the output image.
outputFormat	"png" | "jpeg" | "webp"	"png"	Encoding format for the returned buffers.
password	string	-	Password for encrypted PDFs.
crop	{ x, y, width, height }	-	Crop rectangle in rendered image pixel coordinates.
renderAnnotations	boolean	true	Whether to render annotations and form data.

Return Value: Promise<ConvertedPage[]>

type ConvertedPage = {
  pageIndex: number;
  data: Buffer;
  mimeType: "image/png" | "image/jpeg" | "image/webp";
  width: number;
  height: number;
  dpi: number;
};

⚡ Performance Benchmark

Tested on Apple Silicon (M4). Results represent a local benchmark run against the included fixture PDFs at 300 DPI.

Library	Avg/Page	Relative Speed
pdf-raster	0.86 ms	Baseline (Fastest)
pdfjs-dist + @napi-rs/canvas	5.98 ms	~6.9x slower
pdfjs-dist + node-canvas	7.19 ms	~8.4x slower

Note

Benchmark results are based on the included local fixtures. Performance will vary with PDF complexity and hardware. Run bun run benchmark on your own machine to compare against your environment.

🌍 Runtime & Platform Support

This package uses native bindings and is intended for server-side environments only.

Runtime	Supported
Node.js	✅ Yes
Bun	✅ Yes
Browser	❌ No
Edge Runtimes	❌ No

Supported Architectures:

• OS: Linux (gnu), macOS, Windows
• Arch: x64, arm64

Caution

Do not import this into browser bundles, React Client Components, or Vercel Edge Functions.

🛠 Local Development

This repository is a Bun + Turborepo monorepo.

1. Setup: bun install
2. Native Binaries: bun run pdfium:download
3. Build: bun run build
4. Test: bun run test
5. Benchmark: bun run benchmark

Workspace layout

• core/ - the published pdf-raster package
• example/ - the polished internal demo app using the local workspace package
• consumer/ - a minimal Next.js app for validating the published npm package
• docs/ - the hosted documentation source
• benchmark/ - internal performance comparisons against pdfjs-dist

📄 License

MIT