Wake word detection and speech-to-intent recognition running on a microcontroller. Adds voice command and control to any device with no cloud dependency.
An embedded AI voice assistant listens for a wake word and infers the user's intent directly from the follow-on spoken command, all on a microcontroller with no cloud connection. Porcupine Wake Word and Rhino Speech-to-Intent compose into a lightweight pipeline that fits on ARM Cortex-M devices. Porcupine is always listening with minimal power draw. After the user says the wake word, Rhino processes the follow-on command and extracts a structured intent with slots, ready for the application to act on. No speech-to-text transcription, no NLU parsing, no cloud round-trip.
Porcupine Wake Word provides always-on, low-power wake word detection for embedded devices. It listens continuously on the MCU with minimal CPU usage and triggers Rhino only when the wake word is detected, keeping power consumption low between activations. Custom wake words can be trained in seconds using the Picovoice Console and exported for ARM Cortex-M, Arduino, and all supported platforms.
Rhino Speech-to-Intent infers structured intents directly from spoken commands without an intermediate speech-to-text step, achieving a higher accuracy than Big Tech alternatives, like Google Dialogflow and Amazon Lex. This joint optimization produces a much smaller model than a separate STT + NLU pipeline, which is why Rhino fits on microcontrollers with tight memory constraints. Rhino's models are domain-specific where you define the intents and slots for your product (e.g., a coffee maker, a thermostat, an industrial controller) in the Picovoice Console, and Rhino only recognizes commands within that domain, which improves accuracy and eliminates hallucinations and out-of-domain misrecognition.
Appliance manufacturers can add voice commands directly to ovens, coffee makers, washing machines, and thermostats. The user says "Hey Barista, make a large latte" or "Preheat to 220 degrees," and the device acts immediately. Porcupine and Rhino run on the appliance's embedded MCU without requiring a companion app.
Manufacturers of wearables and hearables can add hands-free voice commands to glasses, helmets, and wearables when users' hands and eyes are occupied. "Check air quality," "Log exercise," "Call supervisor." The pipeline runs within the power and memory budget of battery-operated wearable hardware.
Medical device manufacturers can embed voice control into diagnostic kiosks, infusion pumps, or patient monitors. "Start test," "Pause infusion," "Show vitals." Clinicians interact with equipment by voice while their hands stay with the patient, the chart, or the procedure. All processing stays on the device, simplifying regulatory compliance.
Equipment manufacturers can add voice commands to CNC machines, test instruments, packaging lines, robotic arms, and lab equipment: 'Start cycle,' 'Set speed to 1200,' 'Emergency stop.' Operators keep their hands on the controls while issuing commands. No network infrastructure required on the factory floor. No out-of-domain misrecognition.
Train a wake word and voice commands, flash to your MCU, and run. No cloud required.