Imager::Font - Font handling for Imager.


  $t1font = Imager::Font->new(file => 'pathtofont.pfb');
  $ttfont = Imager::Font->new(file => 'pathtofont.ttf');
  $w32font = Imager::Font->new(face => 'Times New Roman');

  $blue = Imager::Color->new("#0000FF");
  $font = Imager::Font->new(file  => 'pathtofont.ttf',
                            color => $blue,
                            size  => 30);

   $right_bearing) = $font->bounding_box(string=>"Foo");

  my $bbox_object = $font->bounding_box(string=>"Foo");

  # documented in Imager::Draw
  $img->string(font  => $font,
             text  => "Model-XYZ",
             x     => 15,
             y     => 40,
             size  => 40,
             color => $red,
             aa    => 1);


This module handles creating Font objects used by imager. The module also handles querying fonts for sizes and such. If both T1lib and freetype were avaliable at the time of compilation then Imager should be able to work with both truetype fonts and t1 postscript fonts. To check if Imager is t1 or truetype capable you can use something like this:

  use Imager;
  print "Has truetype"      if $Imager::formats{tt};
  print "Has t1 postscript" if $Imager::formats{t1};
  print "Has Win32 fonts"   if $Imager::formats{w32};
  print "Has Freetype2"     if $Imager::formats{ft2};

This creates a font object to pass to functions that take a font argument.

  $font = Imager::Font->new(file  => 'denmark.ttf',
                            index => 0,
                            color => $blue,
                            size  => 30,
                            aa    => 1);

This creates a font which is the truetype font denmark.ttf. It's default color is $blue, default size is 30 pixels and it's rendered antialised by default. Imager can see which type of font a file is by looking at the suffix of the filename for the font. A suffix of 'ttf' is taken to mean a truetype font while a suffix of 'pfb' is taken to mean a t1 postscript font. If Imager cannot tell which type a font is you can tell it explicitly by using the type parameter:

  $t1font = Imager::Font->new(file => 'fruitcase', type => 't1');
  $ttfont = Imager::Font->new(file => 'arglebarf', type => 'tt');

The index parameter is used to select a single face from a font file containing more than one face, for example, from a Macintosh font suitcase or a .dfont file.

If any of the color, size or aa parameters are omitted when calling Imager::Font-new()> the they take the following values:

  color => Imager::Color->new(255, 0, 0, 0);  # this default should be changed
  size  => 15
  aa    => 0
  index => 0

To use Win32 fonts supply the facename of the font:

  $font = Imager::Font->new(face=>'Arial Bold Italic');

There isn't any access to other logical font attributes, but this typically isn't necessary for Win32 TrueType fonts, since you can contruct the full name of the font as above.

Other logical font attributes may be added if there is sufficient demand.


  • file - name of the file to load the font from.

  • face - face name. This is used only under Win32 to create a GDI based font. This is ignored if the file parameter is supplied.

  • type - font driver to use. Currently the permitted values for this are:

    • tt - Freetype 1.x driver. Supports TTF fonts.

    • t1 - T1 Lib driver. Supports Postscript Type 1 fonts. Allows for synthesis of underline, strikethrough and overline.

    • ft2 - Freetype 2.x driver. Supports many different font formats. Also supports the transform() method.

  • color - the default color used with this font. Default: red.

  • size - the default size used with this font. Default: 15.

  • utf8 - if non-zero then text supplied to $img->string(...) and $font->bounding_box(...) is assumed to be UTF 8 encoded by default.

  • align - the default value for the $img->string(...) align parameter. Default: 1.

  • vlayout - the default value for the $img->string(...) vlayout parameter. Default: 0.

  • aa - the default value for the $im->string(...) aa parameter. Default: 0.

  • index - for font file containing multiple fonts this selects which font to use. This is useful for Macintosh DFON (.dfont) and suitcase font files.

    If you want to use a suitcase font you will need to tell Imager to use the FreeType 2.x driver by setting type to 'ft2':

      my $font = Imager::Font->new(file=>$file, index => 1, type=>'ft2')
        or die Imager->errstr;

Returns the bounding box for the specified string. Example:

  my ($neg_width,
      $right_bearing) = $font->bounding_box(string => "A Fool");

  my $bbox_object = $font->bounding_box(string => "A Fool");

the relative start of a the string. In some cases this can be a negative number, in that case the first letter stretches to the left of the starting position that is specified in the string method of the Imager class


how far down the lowest letter of the entire font reaches below the baseline (this is often j).


how wide the string from the starting position is. The total width of the string is $pos_width-$neg_width.


the same as <$global_descent> and <$global_ascent> except that they are only for the characters that appear in the string.


the distance from the start point that the next string output should start at, this is often the same as $pos_width, but can be different if the final character overlaps the right side of its character cell.


The distance from the right side of the final glyph to the end of the advance width. If the final glyph overflows the advance width this value is negative.

Obviously we can stuff all the results into an array just as well:

  @metrics = $font->bounding_box(string => "testing 123");

Note that extra values may be added, so $metrics[-1] isn't supported. It's possible to translate the output by a passing coordinate to the bounding box method:

  @metrics = $font->bounding_box(string => "testing 123", x=>45, y=>34);

This gives the bounding box as if the string had been put down at (x,y) By giving bounding_box 'canon' as a true value it's possible to measure the space needed for the string:

  @metrics = $font->bounding_box(string=>"testing",size=>15,canon=>1);

This returns tha same values in $metrics[0] and $metrics[1], but:

 $bbox[2] - horizontal space taken by glyphs
 $bbox[3] - vertical space taken by glyphs

Returns an Imager::Font::BBox object in scalar context, so you can avoid all those confusing indices. This has methods as named above, with some extra convenience methods.

Parameters are:

  • string - the string to calculate the bounding box for. Required.

  • size - the font size to use. Default: value set in Imager::Font->new(), or 15.

  • sizew - the font width to use. Default to the value of the size parameter.

  • utf8 - For drivers that support it, treat the string as UTF8 encoded. For versions of perl that support Unicode (5.6 and later), this will be enabled automatically if the 'string' parameter is already a UTF8 string. See UTF8 for more information. Default: the utf8 value passed to Imager::Font->new(...) or 0.

  • x, y - offsets applied to @box[0..3] to give you a adjusted bounding box. Ignored in scalar context.

  • canon - if non-zero and the x, y parameters are not supplied, then $pos_width and $global_ascent values will returned as the width and height of the text instead.


The $img->string(...) method is now documented in "string" in Imager::Draw

align(string=>$text, size=>$size, x=>..., y=>..., valign => ..., halign=>...)

Higher level text output - outputs the text aligned as specified around the given point (x,y).

  # "Hello" centered at 100, 100 in the image.
  my ($left, $top, $right, $bottom) = 
                 x=>100, y=>100, 
                 halign=>'center', valign=>'center', 

Takes the same parameters as $font->draw(), and the following extra parameters:


Possible values are:


Point is at the top of the text.


Point is at the bottom of the text.


Point is on the baseline of the text (default.)


Point is vertically centered within the text.


The point is at the left of the text.


The point is at the start point of the text.


The point is horizontally centered within the text.

The point is at the right end of the text.


The point is at the end point of the text.


The image to draw to. Set to undef to avoid drawing but still calculate the bounding box.

Returns a list specifying the bounds of the drawn text.

dpi(xdpi=>$xdpi, ydpi=>$ydpi)

Set or retrieve the spatial resolution of the image in dots per inch. The default is 72 dpi.

This isn't implemented for all font types yet.

Possible parameters are:

  • xdpi, ydpi - set the horizontal and vertical resolution in dots per inch.

  • dpi - set both horizontal and vertical resolution to this value.

Returns a list containing the previous xdpi, ydpi values.


Applies a transformation to the font, where matrix is an array ref of numbers representing a 2 x 3 matrix:

  [  $matrix->[0],  $matrix->[1],  $matrix->[2],
     $matrix->[3],  $matrix->[4],  $matrix->[5]   ]

Not all font types support transformations, these will return false.

It's possible that a driver will disable hinting if you use a transformation, to prevent discontinuities in the transformations. See the end of the test script t/t38ft2font.t for an example.

Currently only the ft2 (Freetype 2.x) driver supports the transform() method.

See samples/ for a sample using this function.

Note that the transformation is done in font co-ordinates where y increases as you move up, not image co-ordinates where y decreases as you move up.


Checks if the characters in $text are defined by the font.

In a list context returns a list of true or false value corresponding to the characters in $text, true if the character is defined, false if not. In scalar context returns a string of NUL or non-NUL characters. Supports UTF8 where the font driver supports UTF8.

Not all fonts support this method (use $font->can("has_chars") to check.)

  • string - string of characters to check for. Required. Must contain at least one character.

  • utf8 - For drivers that support it, treat the string as UTF8 encoded. For versions of perl that support Unicode (5.6 and later), this will be enabled automatically if the 'string' parameter is already a UTF8 string. See UTF8 for more information. Default: the utf8 value passed to Imager::Font->new(...) or 0.


Returns the internal name of the face. Not all font types support this method yet.

glyph_names(string=>$string [, utf8=>$utf8 ][, reliable_only=>0 ] );

Returns a list of glyph names for each of the characters in the string. If the character has no name then undef is returned for the character.

Some font files do not include glyph names, in this case Freetype 2 will not return any names. Freetype 1 can return standard names even if there are no glyph names in the font.

Freetype 2 has an API function that returns true only if the font has "reliable glyph names", unfortunately this always returns false for TTF fonts. This can avoid the check of this API by supplying reliable_only as 0. The consequences of using this on an unknown font may be unpredictable, since the Freetype documentation doesn't say how those name tables are unreliable, or how FT2 handles them.

Both Freetype 1.x and 2.x allow support for glyph names to not be included.


This is used by Imager's string() method to implement drawing text. See "string" in Imager::Draw.


The Freetype 2 driver supports multiple master fonts:


Test if the font is a multiple master font.


Returns a list of the axes that can be changes in the font. Each entry is an array reference which contains:

  1. Name of the axis.

  2. minimum value for this axis.

  3. maximum value for this axis


Blends an interpolated design from the master fonts. @values must contain as many values as there are axes in the font.

For example, to select the minimum value in each axis:

  my @axes = $font->mm_axes;
  my @coords = map $_->[1], @axes;

It's possible other drivers will support multiple master fonts in the future, check if your selected font object supports the is_mm() method using the can() method.


There are 2 ways of rendering Unicode characters with Imager:

  • For versions of perl that support it, use perl's native UTF8 strings. This is the simplest method.

  • Hand build your own UTF8 encoded strings. Only recommended if your version of perl has no UTF8 support.

Imager won't construct characters for you, so if want to output unicode character 00C3 "LATIN CAPITAL LETTER A WITH DIAERESIS", and your font doesn't support it, Imager will not build it from 0041 "LATIN CAPITAL LETTER A" and 0308 "COMBINING DIAERESIS".

To check if a driver supports UTF8 call the utf8 method:


Return true if the font supports UTF8.

Native UTF8 Support

If your version of perl supports UTF8 and the driver supports UTF8, just use the $im->string() method, and it should do the right thing.

Build your own

In this case you need to build your own UTF8 encoded characters.

For example:

 $x = pack("C*", 0xE2, 0x80, 0x90); # character code 0x2010 HYPHEN

You need to be be careful with versions of perl that have UTF8 support, since your string may end up doubly UTF8 encoded.

For example:

 $x = "A\xE2\x80\x90\x41\x{2010}";
 substr($x, -1, 0) = ""; 
 # at this point $x is has the UTF8 flag set, but has 5 characters,
 # none, of which is the constructed UTF8 character

The test script t/t38ft2font.t has a small example of this after the comment:

  # an attempt using emulation of UTF8


If you don't supply a 'type' parameter to Imager::Font->new(), but you do supply a 'file' parameter, Imager will attempt to guess which font driver to used based on the extension of the font file.

Since some formats can be handled by more than one driver, a priority list is used to choose which one should be used, if a given format can be handled by more than one driver.


The current priorities can be retrieved with:

  @drivers = Imager::Font->priorities();

You can set new priorities and save the old priorities with:

  @old = Imager::Font->priorities(@drivers);

If you supply driver names that are not currently supported, they will be ignored.

Imager supports both T1Lib and Freetype2 for working with Type 1 fonts, but currently only T1Lib does any caching, so by default T1Lib is given a higher priority. Since Imager's Freetype2 support can also do font transformations, you may want to give that a higher priority:

  my @old = Imager::Font->priorities(qw(tt ft2 t1));


Arnar M. Hrafnkelsson, And a great deal of help from others - see the README for a complete list.


You need to modify this class to add new font types.

The $pos_width member returned by the bounding_box() method has historically returned different values from different drivers. The Freetype 1.x and 2.x, and the Win32 drivers return the max of the advance width and the right edge of the right-most glyph. The Type 1 driver always returns the right edge of the right-most glyph.

The newer advance_width and right_bearing values allow access to any of the above.


$Revision: 1023 $


Imager(3), Imager::Font::FreeType2(3), Imager::Font::Type1(3), Imager::Font::Win32(3), Imager::Font::Truetype(3), Imager::Font::BBox(3)