A JavaScript (JXA) to perform OCR on images/PDFs using macOS built-in OCR engine

ocr.js

#!/usr/bin/osascript -l JavaScript

ObjC.import("stdlib");
ObjC.import("AppKit");
ObjC.import("PDFKit");
ObjC.import("Vision");

const scriptName = $.NSProcessInfo.processInfo.arguments.objectAtIndex(3).lastPathComponent.js;

console.error = (obj) => {
  const data = $(`${obj}\n`).dataUsingEncoding($.NSUTF8StringEncoding);
  $.NSFileHandle.fileHandleWithStandardError.writeData(data);
};

function ocr(filePath, lang, dpi) {
  const request = $.VNRecognizeTextRequest.alloc.init;
  request.setRecognitionLevel($.VNRequestTextRecognitionLevelAccurate);
  request.setUsesLanguageCorrection = true;

  function ocrPDF(fileURL, dpi) {
    const doc = $.PDFDocument.alloc.initWithURL(fileURL);
    return Array(+doc.pageCount)
      .fill()
      .map((_, i) => {
        const scaleFactor = dpi / (72.0 * $.NSScreen.mainScreen.backingScaleFactor);
        const pdfImageRep = $.NSPDFImageRep.imageRepWithData(doc.pageAtIndex(i).dataRepresentation);
        const originalSize = pdfImageRep.bounds.size;
        const targetRect = $.NSMakeRect(0, 0, originalSize.width * scaleFactor, originalSize.height * scaleFactor);
        const image = $.NSImage.alloc.initWithSize(targetRect.size);
        image.lockFocus;
        $.NSColor.whiteColor.set;
        $.NSBezierPath.fillRect(targetRect);
        pdfImageRep.drawInRect(targetRect);
        image.unlockFocus;
        return ocrTIFF(image.TIFFRepresentation);
      })
      .join("\n");
  }

  function ocrImage(fileURL) {
    const scaleFactor = $.NSScreen.mainScreen.backingScaleFactor;
    const bitmapImageRep = $.NSBitmapImageRep.imageRepWithData($.NSImage.alloc.initWithContentsOfURL(fileURL).TIFFRepresentation);
    const srcSize = $.NSMakeSize(bitmapImageRep.pixelsWide / scaleFactor, bitmapImageRep.pixelsHigh / scaleFactor);
    const srcImage = $.NSImage.alloc.initWithSize(srcSize);
    srcImage.addRepresentation(bitmapImageRep);
    const targetRect = $.NSMakeRect(0, 0, srcSize.width, srcSize.height);
    const newImage = $.NSImage.alloc.initWithSize(targetRect.size);
    newImage.lockFocus;
    $.NSColor.whiteColor.set;
    $.NSBezierPath.fillRect(targetRect);
    srcImage.drawInRect(targetRect);
    newImage.unlockFocus;
    return ocrTIFF(newImage.TIFFRepresentation);
  }

  function ocrTIFF(tiff) {
    const requestHandler = $.VNImageRequestHandler.alloc.initWithDataOptions(tiff, $({}));
    requestHandler.performRequestsError($([request]), null);
    return request.results.js.map((result) => result.topCandidates(1).objectAtIndex(0).string.js).join("\n");
  }

  if (lang === "ja") {
    request.setRecognitionLanguages(["ja", "en"]);
  } else {
    request.setRecognitionLanguages(["en"]);
  }
  const fileURL = $.NSURL.fileURLWithPath(filePath);
  if (fileURL.pathExtension.js == "pdf") {
    return ocrPDF(fileURL, dpi);
  } else {
    return ocrImage(fileURL);
  }
}

function usage() {
  console.log(`Usage: ${scriptName} [--lang <LANG>] [--dpi <VALUE>] <input1> <input2> ...
  
  Input file types: pdf/jpg/png
  Options:
    -h, --help      Show help
    --lang <LANG>   Set OCR language (ja/en) (default: ja)
    --dpi <VALUE>   Set DPI value for PDF rasterization (default: 200)`);
}

function run(args) {
  // parse arguments
  let params = [];
  let lang = "ja";
  let dpi = 200;

  let i = 0;
  while (i < args.length) {
    const opt = args[i];
    switch (opt) {
      case "-h":
      case "--help":
        usage();
        $.exit(0);

      case "--lang":
        if (!args[i + 1]) {
          console.error(`${scriptName}: option requires an argument -- ${opt}`);
          $.exit(1);
        }
        lang = args[i + 1];
        i += 1;
        break;

      case "--dpi":
        if (!args[i + 1]) {
          console.error(`${scriptName}: option requires an argument -- ${opt}`);
          $.exit(1);
        }
        dpi = +args[i + 1];
        i += 1;
        break;

      case "--":
      case "-":
        params = [...params, ...args.slice(i + 1)];
        i = args.length;
        break;

      default:
        if (opt.startsWith("-")) {
          console.error(`${scriptName}: illegal option -- '${opt.replace(/^-*/, "")}'`);
          $.exit(1);
        } else {
          params = [...params, opt];
        }
        break;
    }
    i += 1;
  }

  // handle invalid arguments
  if (!params.length) {
    console.error(`${scriptName}: too few arguments`);
    console.error(`Try '${scriptName} --help' for more information.`);
    $.exit(1);
  }

  for (const file of params) {
    console.log(ocr(file, lang, dpi));
  }
}