Touch, voice or both: choosing interaction modes for a desk pet

Every interaction channel you add to a desk pet brings its own silicon, its own power draw and its own ways to disappoint. The channel list, not the feature list, is where a desk pet project is really scoped — and the honest way to choose is to start from what the product is for, not from what the demo can show.

Start from the core promise

A desk pet sells presence: something alive on the desk that notices you. Presence is carried mostly by reaction — to touch, to being moved, to time of day. Voice conversation is a different promise: a companion that answers. Both are valid products; they just cost different things. Adding conversation to a presence product means mic arrays, wake-word licensing, acoustic echo cancellation tuning and a stance on what happens to the audio — four projects hiding in one checkbox.

Touch: few zones, readable responses

Capacitive sensing under the shell is the cheapest aliveness money can buy, but the design rule is few zones with distinct responses. A pet that reacts identically wherever it is touched trains users to stop exploring. Three zones — head, back, side — each with its own behaviour, produce more perceived interactivity than ten zones sharing one reaction. Zone placement follows the character’s body language, and the sensing layout follows the zones.

Voice: decide the wake story first

If conversation is in the product, the first decision is how it starts listening: a wake word, a button, or always-listening with intent detection. Wake words are familiar but bring licensing and false-trigger tuning. A button is unromantic but predictable. Always-on is the most magical and the most expensive in both power and trust. Each choice reshapes the mic hardware and the firmware schedule, so it belongs at the front of the project, not the middle.

Gesture earns a place as garnish

Gesture sensing demos beautifully and disappoints routinely. Detection depends on distance and lighting, so the same wave that works at the developer’s desk fails under office lighting. As a bonus trigger — wave for a greeting — it delights. As a primary channel it becomes the product’s most common complaint. We recommend it last, if at all.

Combinations are the payoff

The strongest desk pets combine channels into behaviours: a stroke during voice playback changes what it says next; being lifted mid-conversation makes it react to the ride. These cross-channel behaviours cost little incremental hardware and are exactly what users describe as “it actually pays attention”. Plan the channels so at least two can overlap, and the interaction design gets a whole extra layer for free.

If you are choosing channels for a desk pet, we develop these products end to end — sensing, voice, expression and the behaviours between them. The OEM/ODM page outlines the process, or write to us with the interactions your character needs.