Color, personality tone, voice, accessories, and the role the companion plays — all configurable. The interactive configurator demonstrates the full flow.

Color theme

Primary, secondary, and accent palettes drive the shell, LEDs, and on-screen identity.

Personality tone

Calm, playful, professional, motivational — tuned to how the agent speaks and behaves.

Voice style

A placeholder voice profile per character, from soft to professional.

Accessories

Modular snap-on parts — headphones, antenna, visor, badge, and more.

Purpose

The role defines the system prompt, memory profile, and default behavior.

STEM tutorLanguage buddyBrand mascotHome companionReceptionist

Open the configurator

See the live configurator

Demo interaction

A conversation with Kibo

A mock exchange illustrating the voice-agent personality layer. In the prototype, this runs through speech-to-text, an LLM personality agent, and text-to-speech.

YouKibo, can you explain what a sensor does?

KiboOf course! A sensor is how I notice the world — like little senses. A light sensor feels brightness, a touch sensor feels a tap. Want to try one together?

YouYes! Which one should we start with?

KiboLet's start with the light sensor — wave your hand over me and watch my eyes change. Ready when you are.

AI brain, 3D-printed body, simple robotics

AI brain

An LLM personality agent gives each character its voice, tone, and memory — the part that actually holds a conversation.

3D-printed body

A modular printed shell houses the electronics, with replaceable faceplates and accessories per character.

Simple robotics

LEDs for expression, a speaker and microphone for voice, and optional servos for small, lifelike motion.

Architecture

How a spoken sentence becomes a response

The prototype pipeline, from a user's voice to the robot's expression.

User voice

Microphone input

Speech-to-text

Transcription

LLM personality agent

Character + memory

Text-to-speech

Voice synthesis

Robot controller

Speaker · LED · servo

Controller outputs

Speaker — voice playback

LED eyes — expression

Servos — head motion

All processing can run via cloud APIs today, with on-device options on the roadmap.

Technical showcase

The stack behind the prototype

A pragmatic mix of accessible hardware and a modern AI layer.

3D-printable shell design

A modular, printable body with replaceable faceplates and snap-on accessories per character.

Raspberry Pi / mini-PC control layer

The main compute runs the agent orchestration, audio I/O, and connectivity.

ESP32 / Arduino for LEDs & servos

A microcontroller drives expressive LED eyes and small servo movements in real time.

LLM-based personality layer

Each character is a configured agent — system prompt, tone, and guardrails.

Memory & profile configuration

Optional, user-approved memory and per-character profiles shape behavior over time.

Camera & sensor integration

Possible presence, touch, and vision sensors for richer, context-aware interaction.

Video storyboard

The 60-second showcase film

Planned scenes for the demo video — from character selection to the final gallery.

≈ 60 seconds

0:00–0:08

01Character selection

The gallery of eight companions; a hand picks Kibo.

0:08–0:20

02Customization

Colors, accessories, voice, and purpose dialed in on the configurator.

0:20–0:32

033D print

Time-lapse of the modular shell printing layer by layer.

0:32–0:42

04Assembly

Faceplate, LED eyes, speaker, and controller snapping together.

0:42–0:54

05AI voice interaction

Kibo wakes, answers a question, and reacts with light and motion.

0:54–1:00

06Final gallery

The finished companions lined up — and the Inovativi sign-off.

A serious look at AI-plus-robotics, built in-house.

Roboko is how Inovativi explores voice agents, character design, and 3D-printed hardware as one connected system.

Explore the characters Talk to Inovativi

Internal prototype / showcase concept — not a certified consumer robotics product.