Poster of Linux kernelThe best gift for a Linux geek


Section: pdf2djvu manual (1) Updated: 06/12/2010
Local index Up


pdf2djvu - creates DjVu files from PDF files  


pdf2djvu [{-o | --outputoutput-djvu-file] [option...] pdf-file
pdf2djvu {-i | --indirectindex-djvu-file [option...] pdf-file
pdf2djvu {--version | --help | -h}


This program creates a DjVu file from the Portable Document Format file pdf-file.  


pdf2djvu accepts the following options:  

Document type, file names

-o, --output=output-djvu-file

Generate a bundled multi-page document. Write the file into output-djvu-file instead of standard output.

-i, --indirect=index-djvu-file

Generate an indirect multi-page document. Use index-djvu-file as the index file name; put the component files into the same directory. The directory must exist and be writable.


Specifies the naming scheme for page identifiers. Consult the "TEMPLATE LANGUAGE" section for the template language description.

The default template is "p{page:04*}.djvu".

For portability reasons, page identifiers:

• must consist only of lowercase ASCII letters, digits, _, +, - and dot,

• cannot start with a dot,

• cannot contain two consecutive dots,

• must end with the .djvu or the .djv extension.


Equivalent to "--pageid-template=prefix{page:04*}.djvu".


Specifies the template for page titles. Consult the "TEMPLATE LANGUAGE" section for the template language description.

The default is to set no page titles.


Resolution, page size

-d, --dpi=resolution

Specifies the desired resolution to resolution dots per inch. The default is 300 dpi. The allowed range is: 72 ≤ resolution ≤ 6000.


Use MediaBox to determine page size. CropBox is used by default.


Specifies the preferred page size to width pixels × height pixels. The actual page size may be altered in order to respect aspect ratio and DjVu limitations on resolution. (This option takes precedence over -d/--dpi.)


Try to guess native resolution by inspecting embedded images. Use with care.

Image quality

--bg-slices=n+...+n, --bg-slices=n,...,n

Specifies the encoding quality of the IW44 background layer. This option is similar to the -slice option of c44. Consult the c44(1) manual page for details. The default is 72+11+10+10.


Specifies the background subsampling ratio. The default is 3. Valid values are integers between 1 and 12, inclusive.


Try to preserve all the foreground layer colors. This is the default.


Reduce foreground layer colors to the web palette (216 colors). This option is not recommended.


Use GraphicsMagick to reduce number of distinct colors in the foreground layer to n. Valid values are integers between 1 and 4080. This option is not recommended.


Discard any color information from the foreground layer.


Render pages as monochrome bitmaps. With this option, --bg-... and --fg-... options are not respected.


Specifies the aggressiveness of the lossy compression. The default is 0 (lossless). Valid values are integers between 0 and 200, inclusive. This option is similar to the -losslevel option of cjb2; consult the cjb2(1) manual page for details. This option is respected only along with the --monochrome option.


Synonym for --loss-level=100.


Enable font and vector anti-aliasing. This option is not recommended.



Don't extract the metadata.

By default:

• The following entries of the document information dictionary are extracted: Title, Author, Subject, Creator, Producer, CreationDate, ModDate. Timestamps are formatted according to m[blue]RFC 3999m[][1], with date and time components separated by a single space.

The XMP metadata is extracted (or created) and updated accordingly.


Keep the original metadata intact.


Don't extract the document outline.


Make hyperlink borders always visible.

By default, a hyperlink border is visible only when the mouse is over the hyperlink.


Force the specified border color for hyperlinks.

--no-hyperlinks, --hyperlinks=none

Don't extract hyperlinks.


Don't extract the text.


Extract the text. Record the location of every word. This is the default.


Extract the text. Record the location of every line, rather that every word.


Extract no text outside the page boundary.


Don't m[blue]NFKCm[][2]-normalize the text.


Filter the text through the command-line. The provided filter must preserve whitespace, control characters and decimal digits.

This option implies --no-nfkc.

-p, --pages=page-range

Specifies pages to convert. page-range is a comma-separated list of sub-ranges. Each sub-range is either a single page (e.g. 17) or a contiguous range of pages (e.g. 37-42). Pages are numbered from 1.

The default is to convert all pages.



-j, --jobs=n

Use n threads to perform conversion. The default is to use one thread.

-j0, --jobs=0

Determine automatically how many threads to use to perform conversion.

Verbosity, help

-v, --verbose

Display more informational messages while converting the file.

-q, --quiet

Don't display informational messages while converting the file.


Output version information and exit.

-h, --help

Display help and exit.



Details of runtime behaviour with respect to parallelism can be controlled by several environment variables. Please refer to the m[blue]OpenMP API specificationm[][3] for details.



Template syntax

The template language is roughly modelled on the m[blue]Python string formatting syntaxm[][4].

A template is a piece of text which contains fields, surrounded by curly braces {}. Fields are replaced with appropriately formatted values when the template is evaluated. Moreover, {{ is replaced with a single { and }} is replaced with a single }.  

Field syntax

Each field consists of a variable name, optionally followed by a shift, optionally followed by a format specification.

The shift is a signed (i.e. starting with a + or - character) integer.

The format specification consists of a colon, followed by a width specification.

The width specification is a decimal integer defining the minimum field width. If not specified, then the field width will be determined by the content. Preceding the width specification with a zero (0) character enables zero-padding.

The width specification is optionally followed by an asterisk (*) character, which increases the minimum field width to the width of the longest possible content of the variable.  

Available variables

page, spage

Page number in the PDF document.


Page number in the DjVu document.



Layer separation algorithm

Unless the --monochrome option is on, pdf2djvu uses the following naïve layer separation algorithm:

1. For each page, do the following:

1. Raster the page into a pixmap, in the usual manner.

2. Raster the page into another pixmap, omitting the following page elements:

• text,

• 1 bit-per-pixel raster images,

• vector elements (except fills of large areas).

3. Compare both pixmaps, pixel by pixel:

1. If their colors match, classify the pixel as a part of the background layer.

2. Otherwise, classify the pixel as a part of the foreground layer.



If you find a bug in pdf2djvu, please report it at m[blue]the issue trackerm[][5].  


djvu(1), djvudigital(1), csepdjvu(1)  


Jakub Wilk <>



Copyright © 2007, 2008, 2009, 2010 Jakub Wilk


RFC 3999
OpenMP API specification
Python string formatting syntax
the issue tracker



Document type, file names
Resolution, page size
Image quality
Verbosity, help
Template syntax
Field syntax
Available variables
Layer separation algorithm

This document was created by man2html, using the manual pages.
Time: 21:24:37 GMT, April 16, 2011