Lightberry provides a plug‑and‑play software stack that adds multimodal perception and emotional intelligence to existing robots. The platform fuses microphone and camera data to perform real‑time speech recognition, facial expression analysis, and emotion inference, exposing results through ROS‑compatible APIs for autonomous, socially appropriate behavior. It runs on‑device with GPU/NPU acceleration to ensure low latency and privacy.
Funding
Funding not disclosed
Founders
Product
Problem
Many robotic platforms are limited to pre‑programmed motions and lack the ability to interpret auditory and visual cues, preventing them from responding appropriately to human emotions or dynamic environments. This gap reduces the usefulness of robots in service, hospitality, and care settings where natural interaction is essential.
Solution
Lightberry offers a plug‑and‑play software stack that endows existing robots with multimodal perception and emotional intelligence. The platform ingests audio streams and camera feeds, runs real‑time speech recognition, facial expression analysis, and emotion inference, and generates context‑aware spoken responses. Integration is achieved through standard robotics middleware (e.g., ROS 2) and a lightweight SDK, allowing developers to add these capabilities without redesigning hardware. Processed data and emotion states are exposed via secure APIs for downstream decision‑making, enabling robots to act autonomously while maintaining socially appropriate behavior.
Target Audience
The primary customers are robotics manufacturers, system integrators, and developers building service, hospitality, or care robots that require natural human‑robot interaction capabilities.
Features
- Multimodal sensor fusion pipeline that combines microphone arrays and RGB/D cameras for synchronized audio‑visual processing
- Pre‑trained speech‑to‑text and natural‑language generation models optimized for edge deployment, delivering low‑latency conversational responses
- Real‑time facial expression detection and emotion classification engine with customizable affective taxonomies
- ROS‑compatible SDK and API layer for seamless integration into existing robot control stacks and simulation environments
- On‑device inference engine supporting GPU and NPU acceleration to minimize bandwidth and preserve privacy
- Configurable behavior rules that map detected emotions to robot actions (e.g., tone modulation, gesture selection)
- Continuous model update mechanism allowing developers to fine‑tune perception models on proprietary datasets