The FPDF class was modified adding UTF-8 support. Moreover, it embeds only the necessary parts of the fonts that are used in the document, making the file size much smaller than if the whole fonts were embedded. These features were originally developed for the mPDF project, and ported from Ian Back's sFPDF LGPL PHP version.

Before you can use UTF-8, you have to install at least one Unicode font in the font directory (or system font folder). Some free font packages are available for download (extract them into the font folder):

  • DejaVu family: Sans, Sans Condensed, Serif, Serif Condensed, Sans Mono (Supports more than 200 languages)
  • GNU FreeFont family: FreeSans, FreeSerif, FreeMono
  • Indic (ttf-indic-fonts Debian and Ubuntu package) for Bengali, Devanagari, Gujarati, Gurmukhi (including the variants for Punjabi), Kannada, Malayalam, Oriya, Tamil, Telugu, Tibetan
  • AR PL New Sung (firefly): The Open Source Chinese Font (also supports other east Asian languages)
  • Alee (ttf-alee Arch Linux package): General purpose Hangul Truetype fonts that contain Korean syllable and Latin9 (iso8859-15) characters.
  • Fonts-TLWG (formerly ThaiFonts-Scalable)

These fonts are included with this library's installers; see Free Font Pack for FPDF below for more information.

Then, to use a Unicode font in your script, pass True as the fourth parameter of add_font.


This example uses several free fonts to display some Unicode strings. Be sure to install the fonts in the font directory first.

#!/usr/bin/env python
# -*- coding: utf8 -*-

from fpdf import FPDF

pdf = FPDF()

# Add a DejaVu Unicode font (uses UTF-8)
# Supports more than 200 languages. For a coverage status see:
pdf.add_font('DejaVu', '', 'DejaVuSansCondensed.ttf', uni=True)
pdf.set_font('DejaVu', '', 14)

text = u"""
English: Hello World
Greek: Γειά σου κόσμος
Polish: Witaj świecie
Portuguese: Olá mundo
Russian: Здравствуй, Мир
Vietnamese: Xin chào thế giới
Arabic: مرحبا العالم
Hebrew: שלום עולם

for txt in text.split('\n'):
    pdf.write(8, txt)

# Add a Indic Unicode font (uses UTF-8)
# Supports: Bengali, Devanagari, Gujarati, 
#           Gurmukhi (including the variants for Punjabi) 
#           Kannada, Malayalam, Oriya, Tamil, Telugu, Tibetan
pdf.add_font('gargi', '', 'gargi.ttf', uni=True) 
pdf.set_font('gargi', '', 14)
pdf.write(8, u'Hindi: नमस्ते दुनिया')

# Add a AR PL New Sung Unicode font (uses UTF-8)
# The Open Source Chinese Font (also supports other east Asian languages)
pdf.add_font('fireflysung', '', 'fireflysung.ttf', uni=True)
pdf.set_font('fireflysung', '', 14)
pdf.write(8, u'Chinese: 你好世界\n')
pdf.write(8, u'Japanese: こんにちは世界\n')

# Add a Alee Unicode font (uses UTF-8)
# General purpose Hangul truetype fonts that contain Korean syllable 
# and Latin9 (iso8859-15) characters.
pdf.add_font('eunjin', '', 'Eunjin.ttf', uni=True)
pdf.set_font('eunjin', '', 14)
pdf.write(8, u'Korean: 안녕하세요')

# Add a Fonts-TLWG (formerly ThaiFonts-Scalable) (uses UTF-8)
pdf.add_font('waree', '', 'Waree.ttf', uni=True)
pdf.set_font('waree', '', 14)
pdf.write(8, u'Thai: สวัสดีชาวโลก')

# Select a standard font (uses windows-1252)
pdf.set_font('Arial', '', 14)
pdf.write(5, 'This is standard built-in font')

pdf.output("unicode.pdf", 'F')

View the result here: unicode.pdf

Metric Files

FPDF will try to automatically generate metrics (i.e. character widths) about TTF font files to speed up their processing.

Such metrics are stored using the Python Pickle format (.pkl extension), by default in the font directory (ensure read and write permission!). Additional information about the caching mechanism is defined in the add_font reference.

TTF metric files often weigh about 650K, so keep that in mind if you use many TTF fonts and have disk size or memory limitations.

By design, metric files are not imported as they could cause a temporary memory leak if not managed properly (this could be an issue in a webserver environment with many processes or threads, so the current implementation discards metrics when FPDF objects are disposed).

In most circumstances, you will not notice any difference about storing metric files vs. generating them in each run on-the-fly (according basic tests, elapsed time is equivalent; YMMV).

Like the original PHP implementation, this library should work even if it could not store the metric file, and as no source code file is generated at runtime, it should work in restricted environments.

For your convenience, this library collected 96 TTF files in an optional "Free Unicode TrueType Font Pack for FPDF", with useful fonts commonly distributed with GNU/Linux operating systems (see above for a complete description). This pack is included in the Windows installers, or can be downloaded separately (for any operating system).

You could use any TTF font file as long embedding usage is allowed in the licence. If not, a runtime exception will be raised saying: "ERROR - Font file filename.ttf cannot be embedded due to copyright restrictions."