I expected another camera for making immersive video. Lenovo's Project Chronos instead looks closer to a motion-sensing system: a device that observes a person and uses that observation to animate a virtual presence.
The CES concept announcement describes an RGB and time-of-flight camera, a computer inside the device, and an avatar controlled through body movement and expression. It is a concept announcement, so the useful discussion is about the design and the questions it raises, rather than a finished developer workflow.
A stereo video camera and a motion sensor can share some hardware concerns while producing very different outputs. Stereo video preserves two views for a person to watch. A tracking system tries to infer a body, its pose and possibly other aspects of the scene. The resulting data can drive an avatar whose appearance is independent of the camera image.
Depth helps separate some ambiguities in an ordinary picture. An arm pointing toward the camera can look short in image space even though it is extended in the room. Distance information can provide additional evidence about that arrangement. It does not remove occlusion: an arm behind the torso still has to be inferred or treated as uncertain.
What I want to know about a system like this is how much of the observation becomes available to the application. Can a developer obtain joint positions and confidence? Are timestamps provided? Can the camera's coordinate frame be related to another device? A polished avatar demonstration can conceal the answers that determine whether the hardware is useful for other work.
There is also a difference between tracking one person and organizing a shared scene. If two people overlap, the system needs to preserve identity. If somebody leaves and returns, the application needs a policy for reconnecting them to their avatar. Those behaviors are part of the interaction, even when the sensor handles the underlying estimation.
A self-contained device could simplify setup by packaging capture and processing together. It could also create a fixed boundary around which software can be used and which outputs can be accessed. Both possibilities deserve attention before planning a larger system around it.
I would be especially interested in a clear developer sample with a plain visualization of the output. A skeleton, coordinate axes and a confidence display would answer different questions from a cinematic avatar. They would make it easier to understand what the device measures and what the presentation supplies.
The shift is interesting regardless of the final product. A camera can become a source of scene data rather than only a source of pictures. For spatial applications, the quality of that data and the freedom to use it may matter more than the appearance of the demonstration built around it.