Contributing Source Code to libvpx

Revised 2011-01-05

This document will help guide you through contributing code to the WebM Project and libvpx.


In general, most of the on-going work on libvpx concerns speed and quality improvement. A developer typically follows this workflow:

  1. Work on the source code.
  2. Recompile the vpxenc encoder binary.
  3. Perform a video encode job with the new binary.
  4. Evaluate the results, comparing to an identical encode done with a prior version of vpxenc.

Once meaningful improvement is achieved, the developer submits his code revisions to libvpx.

Below we describe in more detail how the core libvpx developers evaluate their own work, so that new contributors can replicate the process and have their code contributions approved and merged more easily.

Required Packages

Become a Contributor

All contributors are welcome. The WebM Project website has details on configuring your machine and creating an account in our code review system, so review the information at to get started.

Subscribe to codec-devel

codec-devel is the mailing list (and Google Group) for developers working on the libvpx library or researching next generation codecs. While anyone may post to codec-devel, subscribing gives you unmoderated access.

Though codec-devel offers a web interface, many subscribers treat it as a simple mailing list and interact solely through their mail clients.

Download Test Clips

  1. Create a directory to hold the clips (e.g., ~/derf_cif on Linux or c:/derf_cif on Windows).
  2. Visit with your web browser.
  3. Download each of the the CIF clips found there, in YUV4MPEG (y4m) format, to the directory you created.

Encoding Only a Few Test Clips

Normally, you won’t want to encode the entire set of test clips, which can be very time-consuming. Instead, consider symlinking only the clips you want to use from their storage directory (e.g. ~/derf_cif) to a working directory (e.g. ~/derf_working), and running your tests there.

Download Test Scripts

Test scripts (and this document) are available in the contributor-guide Git project:

$ git clone
$ cd contributor-guide

Symlink or physically copy the following scripts to somewhere in your $PATH:

Download, Build and Test libvpx

  1. To get started, ensure that your system satisfies the build prerequisites.

  2. Next, clone the libvpx repository. After cloning, you’ll be working in the master branch by default.

    $ git clone
    $ cd libvpx

    Important: Although its not the common case, if your idea would necessitate a change to the VP8 format itself, you’ll want to work in the libvpx experimental branch instead:

    $ git clone
    $ cd libvpx
    $ git checkout -b experimental origin/experimental
  3. Configure and build libvpx. Be sure to run the configure script with --enable-internal-stats, which causes the encoder (vpxenc) to output statistical data as well as a video file. For example:

    $ ./configure --disable-codecs --enable-vp8 --enable-internal-stats --enable-debug --disable-install-docs --log=yes
    $ make

    To learn more about configuration options, do:

    $ ./configure --help

    There are several ways to “install” (make install, etc.), so we’ll leave that up to you. Just ensure that the VP8 encoder (vpxenc) is executable and in your $PATH.

  4. Run some encodes using the provided scripts. Example:

    $ cd <test clip directory>
    $ run_tests 200 500 50 baseline

    In this example, run_tests will encode each y4m file in the current directory at bitrates of 200, 250, 300, 350, 400, 450 and 500 kbps (start at 200, end at 500, 50 kbps increments). It will create a directory baseline and write its output there.

    Output consists of webm video files and stt files containing statistics for each encode. If the test encodes are 2-pass (default), an fpf file (first-pass data file) is also written for each input file, and will be reused on subsequent 2-pass encodes if not deleted.

  5. Make your changes to libvpx, rebuild, and re-run the tests using a different output directory name. Example:

    $ run_tests 200 500 50 mytweak

    Alternatively, change the parameters in the batch encoder script (be) that’s called by run_tests. be in turn calls vpxenc.

  6. When the tests are finished, run to generate an HTML-format report that compares mytweak with baseline.

    $ "*stt" baseline mytweak > mytweak.html
  7. View the report by opening mytweak.html in a web browser.

Submit Your Patches

See the detailed information at