Reachy Eyes
Reachy Eyes is the first third-party hardware product for the Reachy Mini, created by community member Daniel Ritchie of the Brain Wave Collective. It adds programmable, expressive LED eyes to the robot, extending its emotional range and enabling new forms of color-based communication.
📦 GitHub: brainwavecollective/reachy-eyes
🤗 Model: danielritchie/vibe-color-model
🤗 Dataset: danielritchie/cinematic-mood-palette
What It Is​
For a social robot, eyes are one of the most expressive channels available. Reachy Eyes gives the Reachy Mini a way to communicate emotion, mood, and state through color, extending the robot beyond gesture and speech into a visual, ambient dimension.
The product includes physical LED eye hardware, a Python control library, onboard firmware, and an embedded ML model that maps emotional context to eye color automatically.
Key Capabilities​
Color & Light — Support for named colors plus full RGB control. Colors can be set per-eye or both simultaneously, with instant snaps or smooth crossfades. Dynamic parameters such as intensity, energy pulsing, and asymmetry, give the eyes an organic living quality.
Expressive Effects — One-shot effects include blink/wink, acknowledgment pulse, and startle spike. A background behavior engine fires natural randomized blinks (including occasional double-blinks) to keep the robot feeling present without active management.
ML-Driven Emotion (ANIMA) — The onboard ML model accepts a five-dimensional emotional state vector (valence, arousal, dominance, complexity, coherence) and maps it to eye color. Joyful states produce warm, energetic colors; melancholy reads as cool and muted. Designed to connect directly to LLM output or sentiment pipelines. The companion Anima affect engine derives these emotional coordinates from text or conversation.
Character Animations — Built-in animations include weewoo (alternating red/blue, American siren style) and pinpon (rapid triple-pulse, French siren style).
Integration​
The eyes connect via USB and are auto-discovered. The Python package integrates with ROS2, reachy-sdk, or any custom pipeline. Operates headlessly once configured.
The onboard ML model was trained on a curated dataset of color palettes derived from cinematic sources — capturing film's rich vocabulary of color-as-emotion. You can also bypass the model entirely and drive the eyes with any external logic via serial commands.
Full API documentation and DIY build instructions are in the GitHub repository.
Reachy Eyes is an independent community project and is not affiliated with or officially endorsed by Pollen Robotics.