Designing emotion without a screen: motion, sound and timing in companion pets

When a companion pet has no screen, there is nowhere to hide. A screen robot can change its mood with a sprite; a soft body has to say everything with motion, sound and timing. That constraint is not a disadvantage — it is the reason screenless pets feel more alive to some owners than robots with faces. But it means the behaviour model is the product, and it deserves most of the development attention.

The body is the display

Emotion without a screen decomposes into four channels. Speed: a startled pet snaps upright; a sleepy one takes three seconds to raise its head. Rhythm: contentment reads in slow, irregular micro-movements, anxiety in tight repeated ones. Posture: ear angle and head height do the work a smile would. And sound: the difference between a purr and a whimper is a mood change without a single servo moving.

The engineering task is giving each mood a signature across all four channels, so the owner’s brain assembles “happy” or “curious” from a pattern rather than a single cue. One-channel emotion reads as a toy; four-channel consistency reads as a creature.

Sensing decides what it can react to

A screenless pet needs to know it is being touched, held, rocked or covered, and each of those calls for different sensing. Touch arrays under the fabric distinguish a stroke from a press. The accelerometer tells being picked up from being nudged. A microphone hears voice presence without needing to transcribe it. Light sensing separates day from night, and being covered from being held — two states that otherwise look identical.

The reaction model ties sensing to behaviour: what it does when stroked at night differs from what it does when stroked in the morning, and that difference is what makes owners describe the toy as having moods rather than modes.

Timing is the cheapest emotion tool

The most under-used tool in companion design is the pause. A response that starts a beat after the voice reads as listening; one that starts instantly reads as a trigger firing. We tune response delays per interaction: touch replies are fast because touch demands acknowledgement, voice replies wait because conversation expects consideration. None of this costs hardware — it costs tuning time on real units with real people.

Behaviour has to survive repetition

The final test of a screenless pet is week five, when the owner has seen every behaviour a hundred times. A fixed script collapses by then. The behaviour model needs a motion library with variation — the same contentment expressed through slightly different movement combinations — so that patterns feel fresh without ever contradicting the character.

If you are shaping a companion pet around behaviour rather than a display, that is exactly the kind of project we develop. See the OEM/ODM page for how the work is structured, or write to us describing how your character should behave when picked up, stroked and spoken to.