Skip to content
SPECIFICTION
Get listed

GNU Ocrad

gnu.org

Free plan LinuxSelf-hosted
Visit gnu.org
Spec score 7.9 #25 of 40 in OCR software

What it is

GNU Ocrad is a free OCR program and library for Linux and self-hosted use. It reads PNG and PNM images and produces recognized text as 8-bit bytes or UTF-8. Its layout analyzer can separate columns and text blocks on printed pages. Ocrad runs as a standalone console application or as a backend for other programs. Documented options cover image transformations, layout detection, character recognition, text filters, and an OCR results file containing structured recognition details. The software uses ISO 10646 internally, with character-set selection through the --charset option. The manual recommends characters at least 20 pixels high and says 300 dpi scanning usually reaches a suitable size. It also notes sensitivity to character defects and difficulty adding recognition for new characters. The project describes Ocrad mainly as a research project, with algorithms that may change in later releases. Latest releases are available through GNU mirrors, while older and testing versions are hosted on Savannah.

Spec sheet

Purpose
GNU Ocrad is an OCR program and library based on feature extraction.Read 11 Oct 2026 · gnu.org
Free plan
YesRead 23 Sep 2026 · gnu.org
Support
The project page identifies the bug-ocrad mailing list as the appropriate forum for discussing errors.Read 11 Oct 2026 · gnu.org
Algorithm limitation
The manual says Ocrad is sensitive to character defects and that adding recognition for new characters is difficult.Read 11 Oct 2026 · gnu.org
Character sets
The manual says Ocrad uses ISO 10646 internally and lets users select character sets with the --charset option.Read 11 Oct 2026 · gnu.org
Download
The project page directs users to GNU mirrors for the latest release and Savannah for older and testing versions.Read 11 Oct 2026 · gnu.org
Image quality
The manual recommends characters at least 20 pixels high and says scanning at 300 dpi usually produces a suitable character size.Read 11 Oct 2026 · gnu.org
Input and output
It reads PNG or PNM images and produces text in 8-bit byte or UTF-8 format.Read 11 Oct 2026 · gnu.org
Layout analysis
Its layout analyser can separate columns and text blocks on printed pages.Read 11 Oct 2026 · gnu.org
Output formats
The manual documents byte and UTF-8 text output, plus an OCR results file option containing structured recognition details.Read 11 Oct 2026 · gnu.org
Processing options
The manual documents optional image transformations, layout detection, character recognition, and text filters.Read 11 Oct 2026 · gnu.org
Research project
The manual describes Ocrad as mainly a research project and says many algorithms are ad hoc and may change in later releases.Read 11 Oct 2026 · gnu.org
Supported inputs
imageRead 23 Sep 2026 · gnu.org
Usage
Ocrad can run as a standalone console application or as a backend to other programs.Read 11 Oct 2026 · gnu.org

Plans and prices

GNU Ocrad Free No price stated on the pages opened Read 11 Oct 2026 · gnu.org

Every GNU Ocrad plan and price →

Strong points and limits

Strong points

  • Free Linux and self-hosted OCR
  • Reads PNG and PNM images
  • Outputs byte or UTF-8 text
  • Separates printed columns and text blocks
  • Can run as a backend

Limits

  • Sensitive to character defects
  • Adding new character recognition is difficult
  • Algorithms may change in later releases
  • Recommended character height is at least 20 pixels

Who it suits

GNU Ocrad suits Linux or self-hosted users processing PNG or PNM scans from a console or another application. It is relevant to readers comfortable with research-oriented software and image-quality requirements.

Verdict

GNU Ocrad offers command-line OCR with layout analysis, multiple text outputs, and backend use. Its research-project status and sensitivity to image and character defects make recognition behavior an important consideration.

Where Specifiction places it

Alternatives to GNU OcradAll alternatives →