Xiph.org Foundation
- Not-for-profit, volunteer-based organization
- Grew out of an open source project
- Provides royalty- and licence-free multimedia technology
Xiph.org Foundation
We are a standards organization
Xiph.org Foundation
We are a standards organization
- But you can't have standards without reference implementations
Xiph.org Standards
We are a standards organization
- But you can't have standards without reference implementations
- In a heterogeneous market you can't get adoption without good tools
We exist because of patents
- Almost all multimedia compression formats are proprietary, patented, or both
- MPEG licensing is "open" but not free
Why Free standards?
- For the Free desktop we need Free standards
- Free standards allow a commodity market
- no proprietary standard has been successful on the internet
Xiph.org Standards
Layers of multimedia
- Storage and streaming: file system, HTTP and RTP
- Encapsulation and multiplexing: Ogg and RTP payloads
- Codecs for specific compression
- Frameworks: gstreamer wraps all of the above
- Tools for creation and manipulation
Xiph Compression technology
- Vorbis for general-purpose audio
- Theora for general-purpose video
- Speex special-purpose for voice
- FLAC lossless audio compression
Multimedia Formats
Other projects
- Icecast Streaming server application
- XSPF XML playlist interchange
- CMML and Annodex
- Command-line tools and examples
Ogg Container
- Native format for Xiph codecs
- Designed for internet transfer
- Simple, very flexible format
Ogg Container
Features and Bugs
- optimized for streaming: strict interleave, no seek table
- bounded overhead: 0.5% - 2% of bitrate
- provides packet framing, seeking signposts, error detection
Ogg Container
Features and Bugs
- Grouping (multiplexing) allows audio, video to be sent together
- Chaining (concatenation) allows a sequence of Ogg streams to play back seamlessly
- timestamp interpretation up to the codec
Simple Ogg stream
- Codec provides a sequence of packets
- These are packed and/or fragmented into structures called pages
Simple Ogg stream
- Packets are opaque codec data of any length
- Pages can be no more than 64 KB
- Limits both overhead and resource use
Simple Ogg stream
- Begin and End packets/pages are special
- Simple (degenerate) stream is one codec's data
Multiplexed Ogg stream
- Simple Ogg stream is valid on its own...
- But also mixed with others
- Pages are interleaved in temporal order
Complex Ogg stream
- Sequence of multiplexed and simple streams is a valid stream
- Send your whole catalog over one connection!
Ogg Theora
Free video compression format
- Based on the VP3 codec donated by On2
- Patented, but under a free license grant
- Adapted to the Ogg container and extended by Xiph
Ogg Theora
Free video format
- Theora just compresses video
- Combine with Vorbis or Speex for audio
- Combine with MNG or text subtitles
- ...everything but menus
Theora Overview
- arbitrary frame size/aspect
- 'blurry' not 'blocky'
- accurate colour
How does Theora work?
- DCT + block motion compensation
- prediction from previous and key frames
- Huffman coding
- Uniqueness is in the details
Theora Stream structure
- Initial headers configure decoder
- Followed by compressed data
Theora Info header
- compresses planar Y'CrCb (YUV) image data
- 4:2:0 Chroma sampling
- Also supports 4:2:2, 4:4:4
- NTSC and PAL colour spaces defined
Theora Info header
- Rational frame rate and aspect ratio
- encoded image dimensions a multiple of 16
- but cropping allows arbitrary sizes
Theora Comment header
- Same format as Vorbis and Speex
- vendor string identifies the encoder
- unordered list of 'key=value' pairs for rough metadata
Theora Table header
- decoder configuration tables are always specified
- Quantization matrices, Huffman tables, filter coeff
- maximum encoder flexibility
Theora Data packets
- one compressed packet per frame
Theora Data packets
- one compressed packet per frame
- keyframes encode a stand-alone frame
Theora Data packets
- one compressed packet per frame
- keyframes encode a stand-alone frame
- they must occur periodically
Theora Data packets
- one compressed packet per frame
- delta frames encode relative to previous data
Current status
- Bitstream frozen 11 months ago
- 1.0alpha3 encoder will always be supported
- reference implementation is alpha pending proposed API changes, but very stable and definitely appropriate to use
- Decoder doesn't support some non-VP3 features
Current status
- Good support on free operating systems
- Filter set now available for MS Direct Show
- Quicktime still needs to be done
Future work
Experimental implementation by Timothy "Derf" Terriberry
- Decoder complete and apparently stable
- Encoder with new psychovisual rate management
- Supports all spec features
- Optimization and multithreading work
In Summary
- Content creation drives format adoption
- We have a good suite of free codecs, use them by default in your software and when you distribute multimedia