Google Showcases Gemma Translator, an Offline Translation Device Powered by Raspberry Pi 5

Google's Creative Lab has demonstrated the Gemma Translator, a prototype offline translation device that runs the Gemma 4 E2B AI model on a Raspberry Pi 5. The device performs speech-to-speech translation entirely locally, requiring no internet connection. This project showcases the viability of running complex, multi-modal AI workflows—including speech recognition, LLM translation, and speech synthesis—entirely on low-cost, resource-constrained edge hardware. It highlights the potential for private, latency-free translation in remote areas or situations without network access. The device utilizes the Gemma 4 E2B model, which features 5.1B total parameters and 2.3B active parameters, orchestrated via Google's LiteRT-LM framework. Speech transcription and synthesis are handled locally using Moonshine AI models to ensure low-latency processing.