SILK
SILK is an audio codec developed by Skype Limited, and few pieces of software engineering have traveled quite as far as this one. It began as an internal fix to a voice-call quality problem and eventually became part of the foundation for Opus, the Internet standard codec used in everything from web browsers to video conferencing. How does a corporate audio project escape the walls of a single application and reshape how the Internet sounds? That is the thread this documentary follows.
Skype had not always built its own voice technology. Early on, the platform licensed iSAC and iLBC from Global IP Solutions. Those licensed codecs gave way to an in-house replacement called SVOPC, but that was not the end of the road either. SILK was developed on a separate branch, running in parallel with SVOPC for over three years before it was ready to take over. The Consumer Electronics Show in January 2009 was where Skype Limited publicly announced SILK, making it one of the more unusual debuts for audio software. A beta version of Skype 4.0, released on the 7th of January 2009, was the first build to carry SILK in a real product. The final release of Skype 4.0 followed on the 3rd of February.
At its core, SILK is built on linear predictive coding, a technique that models the human vocal tract mathematically to compress speech efficiently. Skype Limited specified that it can operate at sampling frequencies of 8, 12, 16, or 24 kHz, covering a wide range from narrow telephone quality up to wideband audio. Bit rates run from 6 to 40 kbit/s, giving developers room to trade quality against bandwidth. One of the more practically important figures is the algorithmic delay: SILK targets just 25 milliseconds, achieved through a 20 ms frame size combined with a 5 ms look-ahead. For a voice call, that kind of low latency is the difference between a conversation that feels live and one that feels like a radio delay. The reference implementation was written in C, keeping it portable across the hardware landscape of the late 2000s.
On the 3rd of March 2009, just weeks after the codec shipped in Skype, the company announced that SILK would be made available under a royalty-free license to third-party software and hardware developers. The licensing structure has a notable split: the codec itself is open-source and freeware with certain restrictions on use and distribution, while the binary SDK is licensed separately. Early access to the SDK was not automatic. Developers had to apply by providing their name, address, phone number, and a description of their intended use. A later version, 1.0.9, dropped that application requirement, though the license still confined use to internal evaluation and testing; commercial products and distributed software remained outside its scope. On the 6th of July 2009, Skype submitted the first draft of the SILK Speech Codec description to the Internet Engineering Task Force as a candidate for a new Internet wideband audio standard, simultaneously publishing the format's source code.
In September 2010, a hybrid codec then called Harmony was submitted to the IETF. Harmony combined SILK with CELT, a separate codec optimized for music and general audio, and the submission was chosen as the final candidate for the new Internet standard. The IETF published Opus as a proposed standard in September 2012. Skype announced at that point that Opus would be the platform's path forward, in effect stepping aside from SILK as a standalone technology in favor of the broader standard it had helped create. The IETF process had transformed a corporate voice codec into common infrastructure for the Internet.
Steam adopted SILK for its integrated in-game and community voice chat on the 22nd of March 2011, a significant moment for the codec outside its original home. Valve extended that adoption further: on the 14th of April, Team Fortress 2 specifically integrated SILK into its in-game voice chat. The GoldSrc engine platform, updated to SteamPipe, followed on the 29th of January 2013. Zoom has also been found using the codec. Each of those adoptions traces back to the royalty-free licensing announcement from March 2009, which set in motion a much wider deployment than a single messaging platform could have generated alone.
Common questions
What is the SILK audio codec and who developed it?
SILK is an audio compression format and codec developed by Skype Limited, now a Microsoft subsidiary. It was created to replace the SVOPC codec inside Skype and was publicly announced at the Consumer Electronics Show in January 2009.
What audio quality and bitrate does the SILK codec support?
SILK supports sampling frequencies of 8, 12, 16, and 24 kHz and bitrates ranging from 6 to 40 kbit/s. It operates with an algorithmic delay of 25 milliseconds, made up of a 20 ms frame size and a 5 ms look-ahead.
When was SILK first released in Skype?
SILK was first integrated into Skype in version 4.0 Beta 3, released on the 7th of January 2009. The final version of Skype 4.0 followed on the 3rd of February 2009.
How is the SILK codec licensed and is it free to use?
The SILK codec is open-source and available royalty-free but with restrictions on use and distribution. The SDK was initially available only by application; later versions allowed download without application but limited use to internal evaluation and testing, excluding commercial products or software distribution.
What is the connection between SILK and the Opus codec?
SILK is one of the two foundations of Opus, the Internet standard codec. A hybrid codec combining SILK with CELT was submitted to the IETF in September 2010 under the working name Harmony and was published as an IETF proposed standard in September 2012.
What platforms and games have used the SILK codec?
Steam adopted SILK for its voice chat on the 22nd of March 2011. Team Fortress 2 integrated it on the 14th of April that same year, and the GoldSrc platform under SteamPipe adopted it on the 29th of January 2013. Zoom has also been found using SILK.