Seventeen companions here are designed to stay on a desk, a shelf or a workstation rather than be carried, worn or cuddled. Past that they differ considerably: some hold conversations, some respond mainly to touch, and one drives itself around on treads. Sizes run from the palm-sized Elecrow AI Panda up to Vector 2.0.
What Makes a Desk Companion Different
A desk companion stays where you put it and becomes part of the space you work in, which changes what matters about it. Floor-roaming features are irrelevant; footprint, noise and cable routing are not. The AI OTTO L2SA and the Daydream Talk Xiao Zhi are built around conversation. The Energize Lab Eilik works almost entirely through touch zones and an expression library. Vector 2.0 is the exception that moves, using treads on the desk surface itself. They share a location rather than a feature set.
Voice, Touch and Movement at Desk Scale
Interaction at this size relies on the screen more than anywhere else in the catalogue. The AI AEKO carries a 1.54-inch display and a built-in camera on an ESP32-S3 board, with voiceprint recognition described in its product information. The Eilik has no conversational AI at all and still reads as expressive, because its emotion engine drives a large animation library from touch input. The AI Xiaoqian works through voice and tactile interaction in a body of roughly 14 by 6 by 9 cm. More features do not automatically mean a better desk presence.
Desk Space, Power and Noise
Measure before ordering, because product photography flatters these devices. Check the footprint and also the clear area the device needs around it, which is a genuine constraint for anything with articulated movement like the AI OTTO. Decide whether permanent USB power is acceptable or whether you want a battery: the Eilik runs roughly 1.5 hours on a 450 mAh cell over USB-C. If the companion will sit beside you during calls, consider how loud it is and whether sounds can be turned down.
Connectivity and Ongoing Costs
The recurring costs are easy to miss. Vector 2.0 needs an active subscription before its voice commands and cloud features work. Several ESP32-S3 devices require configuration against an external AI service, so the hardware alone does not deliver conversation. Language support is specific rather than universal. If you want a camera on your desk model, the camera-equipped options are collected under AI robots with cameras; for the wider hardware category, see AI robots.