notune/android_transcribe_app

What it solves

This project provides a private, 100% offline speech-to-text solution for Android. It allows users to dictate text into any application or generate real-time subtitles for media playing on their device without sending audio data to the cloud.

How it works

The app acts as a system-level speech-to-text provider or a dedicated voice keyboard. It uses a Rust-based core to run speech recognition models locally on the device via the transcribe.cpp inference engine. It supports various GGUF-formatted models, including Parakeet and Whisper, which can be imported by the user to handle transcription tasks entirely on-device.

Who it’s for

It is designed for privacy-conscious Android users who want high-quality, offline voice typing and captioning capabilities that do not rely on external cloud services like Google's built-in voice typing.

Highlights

• Full Privacy: Operates entirely offline with no internet permissions required.
• Universal Integration: Works as a system speech service, a voice-input popup, or a dedicated keyboard.
• Live Subtitles: Provides real-time, on-device captions for any audio or video playing on the phone.
• Model Flexibility: Supports importing custom GGUF speech models, allowing users to choose the performance/size trade-off that fits their needs.