Project 02 / macOS App

Murmur

Problem: Reading long articles and documents strains your eyes. Existing TTS tools require internet and compromise privacy.

Solution: Built a fully offline voice transcription and text-to-speech app for macOS using Apple Neural Engine. Works in 99+ languages with zero latency.

View on GitHub ↗Features

Tech Stack

SwiftUI

WhisperKit ML

Apple Neural Engine

macOS APIs

Impact

Fully Offline

99+ Languages

Role

Solo developer

(0→1)

Platform

macOS Native

Built with SwiftUI for seamless macOS integration and native performance.

ML Engine

WhisperKit

On-device ML using WhisperKit with Metal acceleration for real-time transcription.

Processing

Apple Neural Engine

Leverages Apple Silicon Neural Engine for ultra-fast local inference.

Languages

99+ Supported

Automatic language detection and support for 99+ languages.

Features / 01

Voice Transcription

Real-time speech-to-text using on-device ML. No internet required.

Text-to-Speech

Natural voice synthesis for reading articles, books, and documents.

Offline First

Fully offline processing with Apple Silicon optimization.

99+ Languages

Support for 99+ languages with automatic language detection.

Apple Neural Engine

Leverages Apple Silicon Neural Engine for ultra-fast inference.

Privacy Focused

All processing happens locally. No data leaves your device.

System Architecture / 02

Audio Input

Microphone / System Audio

→

WhisperKit

On-device ML Processing

→

Text Output

Live Transcription / TTS

Murmur processes all audio locally using Apple Neural Engine. No data is sent to external servers, ensuring complete privacy and zero latency.

Live Demo Concept

murmur_v1.0_process.log

[STATUS]:Listening...
[LANG]:Auto-detected: English (en-US)
[TRANSCRIPT]:"Murmur is the fastest offline transcription app for macOS..."
[ENGINE]:Apple Neural Engine (ANE) - Active
_ processing locally... |