이미지: 원문 영상 캡처 (Google Gemma)
Summary
- Google's organization google-gemma has released a new repository called 'gemma-translator' on GitHub
- It's an on-device voice translator that runs fully offline, powered by Gemma 4 and LiteRT-LM
- It includes a web frontend optimized for small 480x320 screens and Moonshine-based speech synthesis
- 공개 조직
- google-gemma (GitHub)
- 핵심 모델
- Gemma 4 + LiteRT-LM
- 음성 합성
- Moonshine
- 라이선스
- Apache-2.0
- 개발 도구
- Google Antigravity 지원으로 제작
- GitHub 지표
- 포크 82, 스타 648
Google's organization google-gemma has released a new open-source project called 'gemma-translator' on GitHub. The project is an on-device voice translator built on Gemma 4 and LiteRT-LM, and is reportedly designed to work completely offline without any network connection.
According to the repository description, the project was built with support from Google Antigravity and includes a web frontend optimized for small portable displays such as 480x320 resolution screens. The backend consists of an API server using Python's http.server, which communicates with the Gemma model. Text-to-speech (TTS) functionality is handled by Moonshine.
The repository contains folders such as backend, deploy, frontend, and stl, along with a deployment script (deploy-pi.sh) and a model download script (download_model.sh), suggesting it was designed for deployment on small hardware such as Raspberry Pi devices. However, specific supported hardware specifications and the list of supported translation languages were not confirmed in the excerpted material.
The project was released under the Apache-2.0 license and, as of the time of checking, had been forked 82 times and starred 648 times.
