TECH FLOW Svět Androida
← Back to the stream
github.com · picked by Petr Mišák · 54d ago

Experiment with Gemma 4 as an offline translator

Source preview: Experiment with Gemma 4 as an offline translator
AI summary

Gemma Translator is an offline voice translator running on-device without internet requirements, powered by Gemma 4 and LiteRT-LM. The project features a web interface optimized for small displays like Raspberry Pi touchscreens, Moonshine-based speech processing, and a Python API server. Complete setup and deployment documentation is available on GitHub, including configuration for systemd kiosk deployment.

The summary is written by AI from the source; it isn’t the newsroom’s opinion. For details, read the source.
Read source

5 people have already opened the source

Tip author’s note

Recently AI models were used to better understand dolphin speech; now experiments are underway with offline translators. I look forward to a time when AI enables us to communicate better among ourselves, but also with other species. Much like in Star Trek.

AI questions & answers
What are the core hardware requirements for running Gemma Translator?

The project requires a Raspberry Pi 5 with 8GB RAM, a microphone or USB audio interface, speaker or headphone output, and a display with minimum resolution of 480x320 pixels. The operating system must be Linux or macOS with Python 3.10+ and Node.js 18+.

How does the voice interface of Gemma Translator work?

The application captures audio from the microphone, transcribes it using Moonshine STT, translates it with Gemma 4, and plays back the result via TTS in the target language. Two users can communicate with each other while each speaking their own language and hearing the translation.

What are the two available keyboard control modes?

Landscape Mode is designed for a single operator and uses one active lane selected via spacebar. Vertical Mode is for two operators who can independently control each lane with dedicated keys. The choice is saved in localStorage and can be changed anytime in settings.

How is Gemma Translator deployed as a permanent service on Raspberry Pi?

The deploy-pi.sh script automates installation of packages, Python environment setup, UI build, model download, and systemd unit registration. It additionally configures LXDE autostart to launch Chromium in kiosk mode at http://localhost:3000.

Questions and answers are written by AI about the topic, not taken from the source; they aren’t the newsroom’s opinion.
Related from the stream
Mentions