@transkripid/pdf-text-replace

pdf-text-replace

Find and replace text in PDF files while preserving formatting.

Features

• Chainable API mimicking JavaScript's String.replace()
• Supports string and RegExp search patterns
• Preserves font styles, colors, and layout
• Handles FlateDecode compressed streams
• Graceful error handling (returns original buffer on failure)
• Automatic Unicode transliteration (CJK, Cyrillic, accented characters → ASCII)
• Pure TypeScript with minimal dependencies (pako for zlib, any-ascii for transliteration)

Installation

# From npm (when published)
npm install pdf-text-replace

# From local path
npm install /path/to/pdf-text-replace

Usage

import { PDF } from 'pdf-text-replace';
import { readFileSync, writeFileSync } from 'fs';

const input = readFileSync('document.pdf');

const modified = new PDF(input)
  .replace('John Doe', 'Jane Smith')
  .replace('[e-mail]', '[e-mail]')
  .replace(/\d{4}-\d{4}-\d{4}/g, 'XXXX-XXXX-XXXX')
  .toBuffer();

writeFileSync('modified.pdf', modified);

API

new PDF(input: Buffer | Uint8Array)

Create a new PDF instance from a buffer.

.replace(search: string | RegExp, replacement: string): this

Queue a text replacement operation. Returns this for chaining.

• search - String or RegExp pattern to find
• replacement - Text to replace matches with

.toBuffer(): Buffer

Apply all queued replacements and return the modified PDF as a Buffer.

Returns the original buffer unchanged if:

• No matches are found
• An error occurs during processing

How It Works

The library parses PDF content streams (both raw and FlateDecode compressed), finds text operators (Tj, TJ), and performs replacements while:

• Preserving the original font and styling
• Adjusting horizontal scaling (Tz operator) when replacement text has different width
• Rebuilding the PDF with updated stream lengths and xref table

Unicode Support

Replacement text containing Unicode characters is automatically transliterated to ASCII for compatibility with standard PDF fonts (WinAnsiEncoding):

// Chinese → Pinyin
.replace('Author', '银宵')        // Becomes "YinXiao"

// Korean → Romanized  
.replace('Name', '스트레이')      // Becomes "seuteulei"

// Cyrillic → Latin
.replace('Hello', 'Привет')       // Becomes "Privet"

// Accented → Plain ASCII
.replace('Name', 'José García')   // Becomes "Jose Garcia"

This uses any-ascii for transliteration.

Limitations

• Only works with PDFs using WinAnsiEncoding (standard Latin text)
• Complex font encodings (CID, Identity-H) are not supported
• Unicode replacement text is transliterated to ASCII (original Unicode cannot be preserved)
• Text split across multiple operators may not be found
• Scanned/image-based PDFs cannot be modified

License

MIT