Voice-activity detection, speech to text, and text to speech all on RP2040

Adafruit reports that voice processing—detection, speech recognition, and text-to-speech—can now run on an eighty-cent Raspberry Pi RP2350 using five hundred twenty kilobytes of memory. The Moonshine Micro framework, created by Pete Warden, includes a trainable fifty-word command recognizer and a neural network voice engine; a demonstration shows it handling WiFi setup through voice alone. Warden believes local, on-device voice will fundamentally reshape how people interact with physical devices.

Listen to this story

Hear this and more stories in a personalized audio briefing.

Open The Chonkerton