Status: In developmentMobileAI product

Offline AI Voice Translator

Two-way speech translation that runs entirely on the phone

English, Hindi and Tamil speech recognition, translation and speech output in airplane mode, with no internet permission in the strict release build.

Status: Preparing Play Store assets (Nov 2026). · Period: Nov 2026 – present · Source: Private repository.

Screenshots are not published yet. The architecture is shown below.

01Overview

A privacy-first translator for conversations where connectivity or privacy is a concern. All speech and translation models run on the device.

02The problem

Conversation translators usually need the cloud, which fails offline and sends private conversations to servers.

03Who it is for

  • Travellers
  • Field workers
  • Families speaking different languages

04My role

Architecture, model selection and licensing review, Flutter and Kotlin implementation.

05Implemented features

  • Two-way speech translation in English, Hindi and Tamil.
  • Signed, side-loaded language packs.
  • Opt-in, encrypted history.

06Engineering challenges and approaches

Models on a phone

Compact ONNX models, a device benchmark plan and a model licence matrix were done before implementation.

07Outcomes and current status

  • Working app in development.

No usage or revenue numbers are published here unless I can verify them.

08Roadmap (not yet shipped)

  • Google Play release.
  • More languages.