Renders TeXy Math for Github Readmes
Pull request Compare This branch is 4 commits behind leegao:master.
Fetching latest commit…
Cannot retrieve the latest commit at this time.
Failed to load latest commit information.


Renders LaTeX for Github Readmes

Make sure that pdflatex is installed on your system.

readme2tex is a Python script that "texifies" your readme. It takes in Github Markdown and replaces anything enclosed between dollar signs with rendered .

In addition, while other Github TeX renderers tend to give a jumpy look to the compiled text,

readme2tex ensures that inline mathematical expressions are properly aligned with the rest of the text to give a more natural look to the document. For example, this equation is preprocessed so that it lines up at the correct baseline for the text. This is the one salient feature of this package compared to the others out there.


Make sure that you have Python 2.7 or above and pip installed. In addition, you'll need to have the programs latex and dvisvgm on your PATH. In addition, you'll need to pre-install the geometry package in .

To install readme2tex, you'll need to run

sudo pip install readme2tex

To compile and render all of its equations, run

python -m readme2tex --output

If you want to do this automatically for every commit of, you can use the --add-git-hook command once to set up the post-commit hook, like so

git stash --include-untracked
git branch svgs # if this isn't already there

python -m readme2tex --output --branch svgs --usepackage tikz --add-git-hook

# modify

git add
git commit -a -m "updated readme"

git stash pop

and every git commit that touches from now on will allow you to automatically run readme2tex on it, saving you from having to remember how readme2tex works. The caveat is that if you use a GUI to interact with git, things might get a bit wonky. In particular, readme2tex will just assume that you're fine with all of the changes and won't prompt you for verification like it does on the terminal.

You can uninstall the hook by deleting .git/hooks/post-commit. See python -m readme2tex --help for a list of what you can do in readme2tex.


Here's a display level equation

The code that was used to render this equation is just

\frac{n!}{k!(n-k)!} = {n \choose k}

Note: you can escape \$ so that they don't render.

Here's an inline equation.

It is well known that if , then .

The code that was used to render this is:

It is well known that if $ax^2 + bx + c = 0$, then $x = \frac{-b \pm \sqrt{b^2 - 4ac}}{2a}$.

Notice that the equations line up with the baseline of the text, even when the height of these two images are different.

Sometimes, you might run into equations that are bottom-heavy, like . Here, readme2tex can compute the correct offset to align this equation to the baseline of your paragraph of text as well.

Tikz (Courtesy of

Did you notice the picture at the top of this page? That was also generated by . readme2tex is capable of handling Tikz code. For reference, the picture

is given by the tikz code

    \path[coordinate] (0,0)  coordinate(A)
                ++( 60:6cm) coordinate(B)
                ++(-60:6cm) coordinate(C);
    \draw[fill=\couleur!\thedensity] (A) -- (B) -- (C) -- cycle;
    \foreach \x in {1,...,15}{%
        \path[coordinate] coordinate(X) at (A){};
        \path[coordinate] (A) -- (B) coordinate[pos=.15](A)
                            -- (C) coordinate[pos=.15](B)
                            -- (X) coordinate[pos=.15](C);
        \draw[fill=\couleur!\thedensity] (A)--(B)--(C)--cycle;

We can see a few other examples, such as this graphical proof of the Pythagorean Theorem.

How about a few snowflakes?


python -m readme2tex --output []

It will then look for a file called and compile it down to a readable Github-ready document.

In addition, you can specify other arguments to, such as:

  • --readme The raw readme to process. Defaults to
  • --output The processed file. Defaults to
  • --usepackage tikz Addition packages to use during compilation. You can specify this multiple times.
  • --svgdir svgs/ The directory to store the output svgs. The default is svgs/
  • --branch master Experimental Which branch to store the svgs into, the default is just master.
  • --username username Your github username. This is optional, and will try to infer this for you.
  • --project project The current github project. This is also optional.
  • --nocdn Ticking this will use relative paths for the output images. Defaults to False.
  • --htmlize Ticking this will output a md.html file so you can preview what the output looks like. Defaults to False.
  • --valign Ticking this will use the valign trick (detailed below) instead. See the caveats section for tradeoffs.
  • --rerender Ticking this will force a recompilation of all equations even if they are already cached.
  • --bustcache Ticking this will ensure that Github renews its image cache. Github may sometimes take up to an hour for changed images to reappear. This is usually not necessary unless you've made stylistic changes.
  • --add-git-hook Ticking this will generate a post-commit hook for git that runs readme2tex with the rest of the specified arguments after each git commit.

My usual workflow is to create a secondary branch just for the compiled svgs. You can accomplish this via

python -m readme2tex --branch svgs --output

However, be careful with this command, since it will switch over to the svgs branch without any input from you.

Technical Tricks

How can you tell where the baseline of an image is?

By prepending every inline equation with an anchor. During post-processing, we can isolate the anchor, which is fixed at the baseline, and crop it out. It's super clowny, but it does the job.


Github does not allow you to pass in custom style attributes to your images. While this is useful for security purposes, it makes it incredibly difficult to ensure that images will align correctly to the text. readme2tex circumvents this using one of two tricks:

  1. In Chrome, the attribute valign=offset works for img tags as well. This allows us to shift the image directly. Unfortunately, this is not supported within any of the other major browsers, therefore this mode is not enabled by default.
  2. In every (reasonably modern) browser, the align=middle attribute will vertically center an image. However, the definition of the vertical "center" is different. In particular, for Chrome, Firefox, (and probably Safari), that center is the exact middle of the image. For IE and Edge however, the center is about 5 pixels (the height of a lower-case character) above the exact center. Since this looks great for non-IE browsers, and reasonably good on Edge, this is the default rendering method. The trick here is to pad either the top or the bottom of the image with extra spaces until the baseline of the equation is at the center. For most equations, this works great. However, if you have a tall equation, like , you'll notice that there might be a lot of slack vertical spacing between these lines. If this is a deal-breaker for you, you can always try the --valign True mode. For most inline equations, this is usually a non-issue.

How to compile this document

python -m readme2tex --usepackage "tikz" --usepackage "xcolor" --output --branch svgs

and of course

python -m readme2tex --usepackage "tikz" --usepackage "xcolor" --output --branch svgs --add-git-hook