Whisplay HAT+ Transforms Google's Offline Gemma Translator into Compact Pi Device

Google's Gemma Translator, an offline voice-to-text device, has been reimagined on the Raspberry Pi 5 by Jadaie Lin using Whisplay HAT+. The original setup required a separate kiosk panel and multiple peripherals. In contrast, Lin's version integrates everything into a single compact slab with a 640 x 480 DSI touchscreen, two microphones and speakers, and motion sensors.

The device uses the same Gemma 4 E2B model scripts and LiteRT-LM translation engine as the original. Moonshine handles transcription, Gemma performs the translation, and the result is spoken through the onboard speakers. A small web UI in React and Vite displays a real-time chat interface with two discussion lanes.

To set up the device, users need Python 3.10 or newer, Node 18, an 8GB Pi 5, and a one-time model fetch from Hugging Face. After initial setup, the device can operate offline without relying on cloud services. Lin has established a waitlist and Discord channel for those interested in this innovative translator.