--- title: "Extract Text From PDF on Linux Command Line (pdftotext)" description: "Extract text from PDF files on Linux command line using pdftotext. Covers install, layout modes, password-protected PDFs, OCR for scanned docs & batch scripts." date: 2026-07-19 categories: ["linux"] tags: ["pdf","command-line"] --- import Button from "@components/widgets/Button.astro"; import Notice from "@components/widgets/Notice.astro"; import ListCheck from "@components/widgets/ListCheck.astro"; import Accordion from "@components/widgets/Accordion.astro"; import Tabs from "@components/widgets/Tabs.astro"; import Tab from "@components/widgets/Tab.astro"; Extracting text from PDF files on the Linux command line is straightforward with `pdftotext`, part of the `poppler-utils` package. It converts PDF documents to plain text. No GUI, no cloud services, no accounts. Install the package, point it at a file, and get text out the other side. This guide covers install, basic usage, every useful option, handling password-protected and scanned PDFs, alternatives when pdftotext is not enough, batch scripts, and troubleshooting. ## What is poppler-utils? Poppler-utils is a collection of command-line utilities for manipulating PDF files. It is based on the [poppler library](https://poppler.freedesktop.org/), a fork of the xpdf library. The star of this package is `pdftotext`, which converts PDF files to plain text with configurable layout, encoding, and page selection. Other tools in the package handle different PDF tasks: ### Key poppler-utils tools at a glance | Tool | Purpose | |------|---------| | **pdftotext** | Convert PDF to plain text (the focus of this article) | | **pdfinfo** | Print PDF metadata: title, author, page count, etc. | | **pdftohtml** | Convert PDF to HTML | | **pdfimages** | Extract images from a PDF | | **pdfseparate** | Split a PDF into single-page files | | **pdfunite** | [Merge PDF files with pdfunite](https://www.bitdoze.com/pdf-merge-linux-cmd/) into one | If you work with PDFs from the terminal regularly, poppler-utils is one of those packages you install once on every server. It pairs well with other [essential Linux commands](https://www.bitdoze.com/linux-commands/) you use daily. ## How to install poppler-utils on Linux Poppler-utils is in the official repositories of every major Linux distribution. Pick your distro below. ```sh sudo apt install poppler-utils ``` ```sh sudo dnf install poppler-utils ``` ```sh sudo pacman -S poppler ``` ```sh sudo apk add poppler-utils ``` ```sh sudo zypper install poppler-tools ``` If you are still on RHEL/CentOS 7, use `sudo yum install poppler-utils`. On anything newer (RHEL 8+, Fedora, AlmaLinux, Rocky Linux), use `dnf`. ### Verify installation Run: ```sh pdftotext --version ``` Expected output (version will vary by distro): ``` pdftotext version 24.02.0 Copyright 2005-2024 The Poppler Developers - http://poppler.freedesktop.org Copyright 1996-2011 Glyph & Cog, LLC ``` If you get `command not found`, the package is not installed or not in your PATH. Re-run the install command for your distro. Some options require minimum poppler versions: - **`-tsv`** needs poppler ≥ 22.05.0 (Ubuntu 22.10+, Fedora 37+, Arch rolling) - **`-remove-hyphens`** needs poppler >= 26.05.0 (May 2026). Only on rolling distros or manual builds, NOT in Ubuntu 24.04 (ships 24.02.0). Check with `pdftotext --version` before relying on these features. ## How to use pdftotext: basic text extraction ### Basic syntax and first extraction The simplest invocation takes an input PDF and produces a text file: ```sh pdftotext input.pdf output.txt ``` If you omit the output filename, pdftotext uses the same name with a `.txt` extension: ```sh pdftotext input.pdf # Creates input.txt in the current directory ``` Verify it worked: ```sh wc -l output.txt ``` If the line count is zero or near-zero, the PDF is likely scanned/image-based. See the OCR section below. ### Reading from stdin and writing to stdout You can pipe PDF data in and text out using `-` as the filename: ```sh cat input.pdf | pdftotext - - ``` This is useful in scripts and pipelines where you do not want intermediate files: ```sh pdftotext report.pdf - | grep -i "quarterly revenue" ``` ## Customize output format with pdftotext options pdftotext has many options for controlling how text comes out. They fall into a few groups. - **`-layout`**: Preserves the original layout of the PDF, including columns, tables, and spacing. This is my default for multi-column documents. ```sh pdftotext -layout input.pdf output.txt ``` - **`-raw`**: Keeps the original text order but ignores layout positioning. Useful when `-layout` garbles text due to unusual fonts. ```sh pdftotext -raw input.pdf output.txt ``` - **`-fixed number`**: Assumes fixed-pitch (monospace) text with the given character width in points. Useful for PDFs that were generated from fixed-width text. ```sh pdftotext -fixed 8 input.pdf output.txt ``` - **`-colspacing number`**: Controls column detection threshold (default: 0.7). Lower values (e.g., 0.3) detect narrower column gaps; higher values (e.g., 1.5) are more lenient. Adjust when columns are getting merged or split incorrectly. ```sh pdftotext -layout -colspacing 0.3 input.pdf output.txt ``` Extract a specific page range: ```sh # Extract only pages 5 through 12 pdftotext -f 5 -l 12 input.pdf output.txt ``` - **`-f number`**: First page to extract (1-based). - **`-l number`**: Last page to extract. Combine with other options: ```sh pdftotext -f 1 -l 5 -upw "mypass" -layout report.pdf pages1-5.txt ``` These options generate structured output for programmatic processing: - **`-bbox`**: Generates HTML with bounding box coordinates for each word. Useful for positional text analysis. ```sh pdftotext -bbox input.pdf output.html ``` - **`-tsv`**: Outputs tab-separated values with bounding box data per block, line, and word. Columns: `level`, `page_num`, `par_num`, `block_num`, `line_num`, `word_num`, `left`, `top`, `width`, `height`, `conf`, `text`. ```sh pdftotext -tsv input.pdf output.tsv ``` Available in Ubuntu 22.10+, Fedora 37+, Arch rolling. Check your version with `pdftotext --version`. - **`-htmlmeta`**: Generates HTML with PDF metadata (title, author) embedded. ```sh pdftotext -htmlmeta input.pdf output.html ``` - **`-bbox-layout`**: Like `-bbox` but preserves layout structure in the HTML output. - **`-nodiag`**: Discards diagonal text. This is useful for removing watermark text that appears at an angle in the PDF. ```sh pdftotext -nodiag input.pdf output.txt ``` - **`-remove-hyphens all|soft|none`**: Controls how end-of-line hyphens are handled: - `all`: Remove all end-of-line hyphens and merge words (default) - `soft`: Only remove soft hyphens (U+00AD), keep ASCII hyphens - `none`: Keep all hyphens and line breaks ```sh pdftotext -remove-hyphens all input.pdf output.txt pdftotext -remove-hyphens soft input.pdf output.txt ``` This option was added in May 2026. It is NOT in Ubuntu 24.04 (ships 24.02.0). Only available on rolling distros (Arch) or via manual compile. Has no effect in `-raw` or `-layout` mode. - **`-eol unix|dos|mac`**: Sets the end-of-line convention in the output text. Default is the host OS convention. ```sh pdftotext -eol dos input.pdf output.txt ``` - **`-enc encoding`**: Sets the output text encoding. Common values: `UTF-8`, `ISO-8859-1`, `ASCII`, `UCS-2`. ```sh pdftotext -enc UTF-8 input.pdf output.txt ``` - **`-nopgbrk`**: Removes form feed characters (`^L`) between pages. Default behavior inserts them. - **`-q`**: Quiet mode. Suppresses error messages. - **`-r number`**: Sets the resolution in DPI (default: 72). Affects output of some rendering-dependent options. - **`-x number -y number -W number -H number`**: Crop area coordinates. Defines a rectangular region to extract text from. ```sh pdftotext -x 100 -y 100 -W 400 -H 600 input.pdf output.txt ``` - **`-cropbox`**: Uses the crop box instead of the media box (with `-bbox`). You can see the full list of options with: ```sh pdftotext -h ``` ## Extract text from password-protected PDFs Many PDFs have restrictions that prevent copying or printing text. pdftotext can handle two types of passwords. ### Using -opw and -upw options - **`-opw "password"`** — Owner password. Bypasses all restrictions (printing, copying, modifying). Use this when the PDF lets you open it but blocks text extraction. ```sh pdftotext -opw "ownerpass" restricted.pdf output.txt ``` - **`-upw "password"`** — User password. Required to open the file at all. Use this when the PDF prompts for a password on open. ```sh pdftotext -upw "userpass" locked.pdf output.txt ``` If you get exit code 3 or "Permission denied" errors, the PDF has restrictions. Try `-opw` first. ### Decrypting with qpdf before extraction For complex encryption or when pdftotext's password options do not work, decrypt the PDF first with `qpdf`: ```sh # Install qpdf sudo apt install qpdf # Decrypt qpdf --password=secret --decrypt locked.pdf unlocked.pdf # Then extract pdftotext unlocked.pdf output.txt ``` This removes all password protection permanently from `unlocked.pdf`, so handle it accordingly. ## When pdftotext doesn't work: OCR for scanned PDFs If your PDF was created by scanning paper documents, it contains images, not text. `pdftotext` will produce empty or near-empty output. You need OCR (optical character recognition) to add a text layer. This is the number one reason pdftotext "doesn't work" for people. A scanned PDF looks like it has text on screen, but the text is actually a picture of text. ### How to detect a scanned/image-based PDF Quick check: ```sh pdftotext suspect.pdf - | wc -c ``` If the byte count is very low (under 100 bytes for a multi-page document), the PDF is almost certainly scanned. You can also use `pdfinfo`: ```sh pdfinfo suspect.pdf | grep Pages ``` If it shows many pages but pdftotext extracts almost nothing, it is scanned. ### Solution 1: OCRmyPDF + Tesseract OCRmyPDF adds an invisible OCR text layer to scanned PDFs, making them searchable and extractable: ```sh # Install sudo apt install ocrmypdf tesseract-ocr # Add OCR text layer ocrmypdf scanned.pdf searchable.pdf # Now extract text pdftotext searchable.pdf output.txt ``` Verify it worked: ```sh pdftotext searchable.pdf - | wc -c # Should show substantial byte count ``` For non-English PDFs, install the appropriate Tesseract language pack: ```sh # French sudo apt install tesseract-ocr-fra # German sudo apt install tesseract-ocr-deu # Then specify language ocrmypdf -l fra scanned.pdf searchable.pdf ``` ### Solution 2: pdftoppm + tesseract directly If OCRmyPDF is not available or you need more control over the OCR process, convert PDF pages to images first, then run Tesseract: ```sh # Convert PDF pages to PPM images at 300 DPI pdftoppm -r 300 scanned.pdf page # OCR the first page tesseract page-1.ppm output # Output will be in output.txt ``` This approach gives you control over resolution and per-page processing, but you need to loop over multiple pages yourself. ## Alternative command-line tools for PDF text extraction pdftotext handles most cases, but sometimes you need a different tool. Here is when to reach for something else. | Tool | When to use it | Install | |------|---------------|---------| | **mutool draw** | Complex layouts where pdftotext garbles columns | `sudo apt install mupdf-tools` | | **pdfgrep** | Search PDFs without extracting everything | `sudo apt install pdfgrep` | | **xpdf pdftotext** | Tabular data (has `-table` mode poppler lacks) | `sudo apt install xpdf` | | **calibre ebook-convert** | Ebook workflows, different extraction engine | `sudo apt install calibre` | If you prefer a web-based GUI for PDF operations, take a look at [self-hosted PDF manipulation with Stirling PDF](https://www.bitdoze.com/stirling-pdf-self-host-manipulation/) — it runs in Docker and handles extraction, conversion, merging, and more through a browser. ### mutool draw (MuPDF) Sometimes handles complex layouts better than pdftotext: ```sh mutool draw -F txt input.pdf -o output.txt ``` ### pdfgrep — search PDFs without full extraction When you just need to find a string in a PDF, not extract the whole thing: ```sh pdfgrep -n "search term" document.pdf # Search across multiple PDFs pdfgrep -rn "search term" *.pdf ``` ### xpdf pdftotext (with -table mode) xpdf's version of pdftotext has options that poppler's does not, notably `-table` for tabular data and `-simple` for single-column layouts: ```sh sudo apt install xpdf pdftotext -table input.pdf output.txt ``` These are two different binaries with the same name. On Ubuntu, installing xpdf may change which `pdftotext` is in your PATH. Check with `which pdftotext` and `pdftotext --version` to confirm which one you are running. xpdf's version shows "xpdf version" in its output; poppler's shows "poppler" copyright. ### calibre ebook-convert A heavier tool, but useful for ebook-oriented workflows: ```sh ebook-convert input.pdf output.txt ``` ## Batch processing and practical scripts ### Extract all PDFs in a directory ```sh for pdf in *.pdf; do pdftotext -layout "$pdf" "${pdf%.pdf}.txt" done ``` This creates a `.txt` file for every `.pdf` in the current directory. ### Pipe extracted text to grep for searching Search within a PDF without creating intermediate files: ```sh pdftotext document.pdf - | grep -i "keyword" ``` After extracting text, you can [transform text case with sed](https://www.bitdoze.com/sed-change-case/) or do further text processing in your pipeline. ### One-liner to verify extraction succeeded Use this in scripts to check if extraction produced real output: ```sh test $(wc -c < output.txt) -gt 10 && echo "OK" || echo "LIKELY SCANNED PDF" ``` Combine with password and page range: ```sh pdftotext -f 1 -l 5 -upw "mypass" -layout report.pdf pages1-5.txt test $(wc -c < pages1-5.txt) -gt 10 && echo "Extraction OK" || echo "Failed or empty" ``` When batch processing, you may also find it useful to [compare folder contents](https://www.bitdoze.com/compare-folders-content-differences/) to verify all expected output files were created. ## Troubleshooting common pdftotext issues **Symptom:** `pdftotext` produces an empty or near-empty output file. **Cause:** The PDF is scanned/image-based with no text layer. **Fix:** Use OCRmyPDF to add an OCR text layer (see the [OCR section](#when-pdftotext-doesnt-work-ocr-for-scanned-pdfs) above). **Symptom:** Extracted text has wrong characters, missing letters, or garbled output. **Cause:** The PDF uses custom font encodings that pdftotext cannot map to Unicode. **Fixes:** 1. Try `-raw` mode to bypass layout processing: ```sh pdftotext -raw input.pdf output.txt ``` 2. Force UTF-8 encoding: ```sh pdftotext -enc UTF-8 input.pdf output.txt ``` 3. If poppler cannot handle it, try `mutool draw -F txt` as an alternative. **Symptom:** Columns are merged together or text from different columns is interleaved. **Cause:** The column detection heuristic does not match the PDF's layout. **Fix:** Adjust `-colspacing` (default: 0.7). Lower values detect narrower column gaps: ```sh pdftotext -layout -colspacing 0.3 input.pdf output.txt ``` Try values between 0.3 and 1.5 until the layout looks right. **Symptom:** Diagonal watermark text (e.g., "CONFIDENTIAL", "DRAFT") appears in the extracted text. **Cause:** pdftotext extracts all text, including diagonal overlays. **Fix:** Use `-nodiag` to discard diagonal text (poppler ≥ 0.80.0): ```sh pdftotext -nodiag input.pdf output.txt ``` For scripting and automation, pdftotext returns meaningful exit codes: | Exit code | Meaning | |-----------|---------| | 0 | Success | | 1 | Error opening PDF file | | 2 | Error opening output file | | 3 | Permission error (PDF has restrictions) | | 99 | Other error | Use in scripts: ```sh pdftotext input.pdf output.txt case $? in 0) echo "Success" ;; 1) echo "Cannot open PDF" ;; 3) echo "PDF has restrictions — try -opw option" ;; *) echo "Unknown error" ;; esac ``` ## Conclusion `pdftotext` from `poppler-utils` is the default tool for extracting text from PDFs on Linux. It handles most cases out of the box: basic conversion, layout preservation, page selection, encoding control, and password-protected files. For scanned PDFs, pair it with OCRmyPDF and Tesseract. For tabular data that pdftotext mangles, try xpdf's `-table` mode or `mutool draw`. For more PDF work on the command line, you can [merge PDF files with pdfunite](https://www.bitdoze.com/pdf-merge-linux-cmd/) or [self-host Stirling PDF](https://www.bitdoze.com/stirling-pdf-self-host-manipulation/) for a browser-based toolkit. And if you are building out your Linux admin skills, check out guides on [securing your SSH server](https://www.bitdoze.com/secure-ssh-server-linux/) and [checking remote ports with nc](https://www.bitdoze.com/check-remote-port-in-linux-nc/).