Documentation

Opus API reference (Doxygen)

libopus 1.0.2 release HTML, PDF
libopus development version HTML, PDF

Opusfile API reference (doxygen)

libopusfile 0.2 release HTML, PDF
libopusfile development version HTML, PDF

Opus-tools man-pages

opusenc (.wav to .opus) HTML
opusdec (.opus to .wav) HTML
opusinfo (.opus file info) HTML

IETF specifications

Official Opus codec RFC
(rfc6716)
HTML, PDF, TXT
Latest Opus RTP draft
(draft-spittka-payload-rtp-opus)
HTML, PDF, TXT
Latest OggOpus container draft
(draft-ietf-codec-oggopus)
HTML, PDF, TXT

Tutorials

Encoding Opus from foobar2000 Windows

Test Vectors

Test files to verify that the Opus decoders are operating properly.

Bitstream conformance set for the codec.
This set of bitstreams is used to measure codec implementations for conformance with the specification.
Also available from Xiph.Org and Mozilla.

The .opus test set is an in-development larger test suite for .opus files.
These are suitable for testing opusdec and other media-player software.
Files prefixed with 'broken' and 'failure' are intentionally malformed and at least the latter should be rejected.

Note that these files are designed to exercise all aspects of an implementation, the fact that they sound terrible is an intended consequence.

Inside Opus

Beyond the RFC itself, there are a several documents out there that describe Opus or parts of Opus. Not everything in them applies to the final version of Opus, but they should still be useful to those who want to understand how Opus works. For an overview of the codec, see the linux.conf.au 2012 Opus presentation. For historical reasons, most of the information on the algorithmic details is specific to either SILK or CELT.

SILK

For information on SILK, see the slides by Koen Vos from the 79th IETF meeting. See also the original SILK codec page.

CELT

The most accessible descriptions of CELT are probaly Monty's CELT demo. You can also see Tim Terriberry's linux.conf.au 2009 CELT presentation. Those with a signal processing background can also read two papers that were published on CELT:

  1. J.-M. Valin, T. B. Terriberry, C. Montgomery, G. Maxwell, A High-Quality Speech and Audio Codec With Less Than 10 ms delay, IEEE Transactions on Audio, Speech and Language Processing, Vol. 18, No. 1, pp. 58-67, 2010.
  2. J.-M. Valin, T. B. Terriberry, G. Maxwell, A Full-Bandwidth Audio Codec with Low Complexity and Very Low Delay, Proc. EUSIPCO, 2009.

Paper 1 describes CELT version 0.3.2, while paper 2 applies to 0.5.2. The main things that changed since these papers were written are the pitch prediction/filtering and the birdie avoidance method.