Artificial intelligence is rapidly transforming the hybrid meeting experience. Features such as intelligent speaker tracking, automatic framing, multi-camera switching, and participant recognition are making meetings more engaging and productive. Yet while AI capabilities continue to evolve, many meeting room architectures were designed around connectivity requirements that predate today’s AI-driven workflows.
USB has become the standard interface for conferencing cameras for good reason: it’s simple, widely supported, and deeply integrated into virtually every conferencing platform. However, USB’s native cable length is limited to just a few meters, creating challenges when cameras need to be placed throughout medium and large meeting spaces.
To overcome these distance limitations, some system designers have replaced USB cameras with IP-based cameras. But since Room PCs and video conferencing appliances don’t have IP video inputs, designers are forced to add another component to convert IP to USB. This approach enables flexibility in room design, but introduces its own challenges: IP requires compression, which affects image quality; it introduces latency, which hinders the conferencing experience; and it isn’t plug-and-play, requiring IT network configuration.
As AI becomes more central to the meeting experience, these architectural decisions take on greater importance. Modern AI algorithms benefit from access to the highest-quality video possible – when a wide shot is cropped and scaled to frame each participant individually, a compressed source image can significantly affect the quality of the final result.
At the same time, conferencing platforms are evolving. Zoom (Intelligent Director and Multi-Stream) and Microsoft Teams Rooms both leverage multiple camera streams to create individual views of in-room participants, reflecting the industry’s shift toward centralised AI processing. Video bar technology is advancing too – Logitech, for example, now offers a product supporting up to five simultaneous video streams.
In short, rather than embedding intelligence in every camera, AI processing can increasingly happen where all video streams converge — opening the door to simpler, more efficient system architectures.
The HDBaseT-USB3 specification was developed with this shift in mind. By extending native USB 3 connectivity over long distances without converting traffic to IP or compressing the video stream, it enables cameras to deliver uncompressed video directly to centralised AI processing with minimal added latency.
The first chipset compliant with the HDBaseT-USB3 specification, Valens Semiconductor’s VS6320, allows manufacturers to build systems that simplify room infrastructure while preserving native USB data from camera to Room PC. Fewer conversion stages mean less hardware and ensure AI applications receive the highest-quality source data available.
For camera manufacturers, the recently introduced Valens VS6310 extends these benefits further. This low-power device can be integrated directly into conferencing cameras, eliminating the need for external extender hardware and enabling cleaner, more streamlined room designs. Operating at 2.5Gbps, the VS6310 supports the bandwidth typical video conferencing cameras require.
As hybrid meetings continue to evolve, advances in AI will depend not only on smarter algorithms but also on the infrastructure that delivers data to them. By enabling native, uncompressed USB connectivity over long distances, the industry can design meeting spaces that are simpler, more efficient, and better equipped to support the next generation of AI-powered collaboration.
You can subscribe to Installation magazine for free here and the daily newsletter here.
