Google Creative Lab builds an offline pocket translator on Raspberry Pi 5
Google Creative Lab built a portable voice translator running locally on a Raspberry Pi 5 using the Gemma 4 language model and Moonshine speech tools.
The Google Creative Lab team has introduced Gemma Translator, a portable voice translator designed to run entirely offline. The device is built around a Raspberry Pi 5 single-board computer equipped with 8 GB of RAM.
Text translation is handled by a compact language model, Gemma 4 E2B, running locally via the LiteRT-LM inference framework. The model's weights require approximately 1.1 GB of memory to load. Speech recognition and synthesis are managed by Moonshine. To use the device, a user holds down a button while speaking; the system then recognizes the phrase, translates it through Gemma, and plays the translated audio in the chosen language.
The interface is adapted for a 320×480 pixel screen styled like a retro terminal, split into two sections so each speaker can select their language. The source code has been published under the Apache 2.0 license alongside deployment scripts and STL files for 3D printing a custom case. The project description notes that Gemma Translator was developed using Google Antigravity.
Source: habr
Comments
Help shape the next version: Add context or suggest a correction. AI review can add points toward a rewrite. Reviews and updates may take time; a full meter does not guarantee a new version.