PDF ocr + compress

0.ocr-pdf.md

OCR PDF

Just a Deno script to run tesseract and ghostscript to produce small pdf with text.

You can use picture or pdf as infiles whatever supported by Tesseract and GhostScript.

Usage

deno run --allow-read \
  --allow-write \
  --allow-run=tesseract,gs \
  https://gist.githubusercontent.com/lpenaud/dbcf1877a8dd2138f9ac64f4a01a4646/raw/ebook.ts \
  input.tiff \
  output.pdf

Dependencies

• tesseract
• ghostscript
• deno

ebook.ts

#!/usr/bin/env -S deno run --allow-read --allow-write --allow-run=tesseract,gs

import * as stdPath from "jsr:@std/path";

class VerboseCommand extends Deno.Command {
  #command: string[];

  constructor(command: string | URL, options: Deno.CommandOptions) {
    super(command, options);
    this.#command = [typeof command === "string" ? command : command.href];
    if (options.args !== undefined) {
      this.#command.push(...options.args);
    }
  }

  override output(): Promise<Deno.CommandOutput> {
    this.#log();
    return super.output();
  }

  override outputSync(): Deno.CommandOutput {
    this.#log();
    return super.outputSync();
  }

  override spawn(): Deno.ChildProcess {
    this.#log();
    return super.spawn();
  }

  #log() {
    console.log(...this.#command);
  }
}

interface GhostScriptOptions {
  infile: string;
  outfile: string;
}

// gs -sDEVICE=pdfwrite -dCompatibilityLevel=1.4 "-dPDFSETTINGS=/ebook" -dNOPAUSE -dQUIET -dBATCH -sOutputFile=final.pdf output.pdf
class GhostScriptCommand extends VerboseCommand {
  constructor({ infile, outfile }: GhostScriptOptions) {
    super("gs", {
      args: [
        "-sDEVICE=pdfwrite",
        "-dCompatibilityLevel=1.4",
        // https://ghostscript.readthedocs.io/en/latest/VectorDevices.html#controls-and-features-specific-to-postscript-and-pdf-input
        // /screen,/printer,/prepress,/default
        "-dPDFSETTINGS=/ebook",
        "-dNOPAUSE",
        "-dQUIET",
        "-dBATCH",
        `-sOutputFile=${stdPath.resolve(outfile)}`,
        stdPath.resolve(infile),
      ],
    });
  }
}

interface TesseractOptions {
  language: string;
  infile: string;
  outfile: string;
}

class TesseractCommand extends VerboseCommand {
  constructor({ infile, language, outfile }: TesseractOptions) {
    const { dir, name } = stdPath.parse(outfile);
    super("tesseract", {
      args: [
        "-l",
        language,
        stdPath.resolve(infile),
        stdPath.resolve(dir, name),
        "pdf",
      ],
    });
  }
}

interface CleanOptions {
  temp: string;
}

function clean({ temp }: CleanOptions) {
  Deno.removeSync(temp);
  console.log("Removed:", temp);
}

function main(args: string[]): number {
  if (args.length !== 2) {
    console.log(
      "Usage:",
      import.meta.filename ?? import.meta.url,
      "INFILE",
      "OUTFILE",
    );
    return 1;
  }
  const [infile, outfile] = args;
  const temp = Deno.makeTempFileSync({
    prefix: "ebook",
    suffix: ".pdf",
  });
  const tesseract = new TesseractCommand({
    infile,
    language: "fra",
    outfile: temp,
  });
  const tesseractOutput = tesseract.outputSync();
  if (!tesseractOutput.success) {
    clean({ temp });
    return tesseractOutput.code;
  }
  const gs = new GhostScriptCommand({
    infile: temp,
    outfile,
  });
  const gsOutput = gs.outputSync();
  clean({ temp });
  return gsOutput.code;
}

if (import.meta.main) {
  Deno.exit(main(Deno.args.slice()));
}