Startup Vidyo took the stage at DEMO today to demonstrate a new approach to videoconferencing, following the announcement of their new offering and a licensing relationship with Cisco Systems.
The Hackensack, N.J.-based startup has implemented a newly minted standard, called scalable video coding (SVC), that enables video streaming at multiple resolutions, quality levels and bit rates. H.264/SVC, as the standard is known, fundamentally changes the way videoconferencing is delivered. And while the standard is public, implementing it is what makes Vidyo special.
“The standard allows you to have layers of encoded video, and if you’re intelligent about how you do that you can do lots of things,” Marty Hollander, SVP of marketing, told me.
To understand the kinds of intelligent things Vidyo might do, we need to look at how videoconferences usually work. Today’s videoconferencing relies on fairly unintelligent devices at the edge, and intelligence in the core — what Hollander calls a “mainframe” model. A multi-point conferencing unit receives streams of video from each participant, decodes them, combines them and re-encodes them.
At DEMO, the company showed live, four-way web-based conferencing (always a risky demo) in which each participant had different local settings — from a high-definition video at 60 frames per second to a consumer-grade VGA connection to a notebook camera. Because of the way they have implemented the new standard, the result was impressive.
The traditional approach of doing encoding and decoding in the core introduces delay. It also means that users all get the same data stream, rather than one suited to their bandwidth, window size and client. In short, transcoding in the cloud makes for lousy conversations.
Vidyo’s implementation fixes this. Their architecture puts more intelligence at the edge and a specialized router in the core. In their model, the router and each participant’s client work together. They tailor data transmission to each client’s demands. When a client shrinks the size of their video window, the router knows — and as a result, sends fewer bits, less frames, or lower-quality video to that client without affecting everyone else. The router isn’t busy encoding, merging and decoding streams, but it knows about processing, window size and bandwidth, which it can adjust dynamically throughout the call.
“The standard tells you how to decode. That’s the easy part. How to encode is harder — that’s the secret sauce,” said Hollander. “We implemented the standard by thinking about how a network would do it.” The author of H.264/SVC, Thomas Weigand, is an adviser to the company.
Vidyo has 40 employees, a $12 million series B financing, and 24 patents pending on the technology.
The router is a 1RU, Linux-based quad-core computer that handles up to 100-way conferences. It’s sold as a subscription model for $1,000 per HD port per year, which is cheaper than the traditionally capital-intensive spending around videoconferencing. Vidyo also makes a video gateway that lets “legacy” videoconferencing hardware work with the new standard.
The video router technology is where the money is, which may be why Cisco announced plans to use Vidyo’s technology into their Unified Communications platforms. The company is cagey on details of the deal, however, since Cisco is licensing the technology itself as opposed to selling Vidyo’s products, and already sells traditional videoconferencing products. But Vidyo did say that Cisco plans on building new products based on Vidyo’s technology in the future. Tal Shalom, entrepreneur in residence with the firm, explained that they work at the application layer. Update: So in theory, the Vidyo system could run on a CPU in a Cisco router (we had previously reported, incorrectly, that the system would run on a board in Cisco’s router.)