whisper.rn

Why we included this project

Mobile apps that need speech-to-text usually end up sending audio to a cloud API, which costs money per minute and raises privacy questions. whisper.rn sidesteps that by running Whisper transcription on the device itself: it wraps whisper.cpp, so inference happens locally on iOS and Android and the audio never leaves the phone. It does more than transcribe recorded files. Live microphone input works with Silero voice activity detection, and direct PCM processing through JSI lets you feed audio straight into the model, which covers voice notes, real-time captions, meeting recordings, and voice-command interfaces. Core ML and Metal acceleration on iOS plus ARM NEON on Android keep it fast enough for real use on phones rather than just demos. If you ship a React Native app that needs offline or low-latency transcription, this is a solid place to start.