This document provides a quick overview over the various ways to customize pandoc's output, with links to fuller documentation and some examples.
--standalone option is used,
pandoc will generate a standalone document rather than a fragment. For
example, in HTML output this will include the
element; in LaTeX output, it will include the preamble.
Pandoc comes with a default template for (almost) every output
format. A template is a plain text file containing variables that are
replaced by text generated by pandoc. For example, the variable
$body$ will be replaced by the document body, and
$title$ by the title from metadata.
To look at the default template for an output format, you can do
pandoc -D FORMAT, where
FORMAT is replaced by
the name of the format. For example
pandoc -D latex. You
can also use your own template instead, either by using the
--template option or by putting the custom template in your
user data directory (on Linux and macOS,
Note that in many cases you can avoid the need for a custom template
by including a file with the
option. Or you can set the corresponding template variable directly.
There are several ways to set template variables:
|values can be…||strings and bools||strings and bools||also YAML objects and lists|
|strings are…||inserted verbatim||escaped||interpreted as markdown|
|accessible by filters:||no||yes||yes|
For more information, see Templates in the pandoc manual.
documents, things are a bit more complicated. Instead of a single
template file, you need to provide a customized
reference.docx/pptx/odt. See the manual for the
Templates are very powerful, but they are only a sort of scaffold to place your document's body text in. You cannot directly change the body text using the template.
If you need to affect the output of the actual body text, you can use a pandoc filter. A filter is a small program that transforms the document, between the parsing and the writing phase, while it is still in pandoc's native format. For example, a filter might find all the Header elements of a document and capitalize their text.
Pandoc's native representation of a document is an abstract syntax
tree (AST), not unlike the HTML DOM. It is documented here.
Pandoc document is a chunk of metadata
Meta) and a list of
Blocks, in turn, are composed of other
Inline elements. (
Block elements are
things like paragraphs, lists, headers, and code blocks.
Inline elements are individual words, links, emphasis, and
so on.) Filters operate on these elements. You can use
pandoc -t native to learn about the AST's structure.
There are two kinds of filters: JSON filters (which transform a JSON serialization of the pandoc AST, and may be written in any language that can parse and emit JSON), and Lua filters (which use an interface built directly into pandoc, and must be written in the Lua language). If you are writing your own filters, it is best to use Lua filters, which are more portable (they require only pandoc itself) and more efficient. See Lua filters for documentation and examples. If you would prefer to write your filter in another language, see Filters for a gentle introduction to JSON filters.
There's a repository of lua filters at pandoc/lua-filters on
GitHub. A number of pandoc filters, written in Haskell, are available on
and can be installed using the
tools. The wiki also lists third party
TODO Divs and Spans: generic blocks that can be transformed with filters
Custom Styles in Docx
TODO Generic raw attributes: to include raw snippets
TODO Custom writers
TODO Custom syntax highlighting, provided by the skylighting library
including highlighting styles