Matroska and Web Streaming

W3C - 8th February 2011

Steve Lhomme

Matroska Chairman

www.matroska.org

Summary

• What is Matroska?
• Matroska and WebM
• Hardware/Software Support
• File Streaming
• Live Streaming
• New features for the Web
• Standardization

What is Matroska ?

• Audio/Video interleaved container
• Developped with technical inputs from many people online
• Completely open and royalty-free
• Initial specifications 8 years old
• Based on EBML, a binary XML (extensible and space efficient)
• More feature packed and efficient than MP4, AVI, MPEG TS, Ogg
• Used a lot for HD and subtitled content
• OSS Sources for about any language available (incl. JavaScript)

Matroska and WebM

WebM is a web specific profile of Matroska

Like XHML Strict and Transitional (header change)

• Allows only VP8 (video) and Vorbis (audio) codec
• Missing chapters, metadata/tags, attachments, signature/encryption
• Can peek mature features of Matroska at any time

Hardware/Software support

.mkv playback is widespread in the industry

Usually H.264/DivX with AC-3/DTS/AAC/Vorbis/MP3

We have a validation tool for better compliancy

• In all computer video players internally or via free plugins
• WebM in Chrome, FireFox, Opera
• In connected TVs and Bluray players (Samsung, Sony, etc)
• In home media streamers (PopCorn Hour, WD TV, Roku, etc)
• In some cellphones (DivX certified or WebM in Android 2.3)
• Editing in open-source video editors

Stream Modes

2 modes read using the same parsers

• Fixed streams with a known end (MOV, AVI, MPG)
• Continuous streams with an unknown end (MPEG TS, Ogg)

Fixed files (HTTP, FTP or local)

One of the design goals was to make a format suitable for file streaming

• Very small header required before playback can start (< 1 KB)
• With the index at the front (< 25 KB for full movie) or No index
• Similar or smaller size as 3GPP/MP4 on codecs they support
• Much lower overhead than Ogg, MPEG TS, AVI
• Error resilient in case of transmission error
• Extensive Tags help the semantic web (like EXIF data on photo sites)

Live Streaming

Unlike XML it can have opened elements with no known end (infinite size)

<cluster>data... </cluster> <cluster>

Allows sending data before the end of the 'chunk' is known

Perfect for low latency streaming

Demonstrated with WebM by Flumotion at GUADEC in July 2010

Independent chunks of data allow for multicasting

Chain segments for streaming reconfiguration (16x9 to 4x3)

New features for the Web

The video world is moving to the web and at the same time adding new technologies

• Adaptive Streaming: key to mobile video & fluid on desktop
• Live Streaming: needs an adequate container (MPEG TS is feature limited and bandwidth hungry)
• Stereoscopic 3D: needs markers in the container (YouTube + WebM + NVIDIA + FireFox 4)
• DRM: remains intact after 3rd party remuxing
• Transparency: for better integration in the page design (planned for WebM)

Standardization

There is already an IETF like RFC for EBML

Already a de facto standard with all the industry support

We need a more formal standardization of the specifications endorsed by a recognized organization for broader industry use

The current specifications and specification notes should cover each field in detail and leave no room for misinterpretation

All work now supervised by a non profit organisation based in France

Questions