Overview:
VoiceInk is an open-source, native macOS application designed for near-instantaneous voice-to-text transcription. It uses local AI models to process speech entirely offline, aiming to keep user data private by never leaving the device. The app is positioned as a privacy-focused alternative for users who need fast and accurate dictation, offering features like configurable shortcuts and context-aware settings that adapt to the active application.
Core Features:
Accurate Local Transcription: Relies on local AI models (based on
whisper.cppand FluidAudio) to transcribe speech to text with high accuracy, processing almost instantly.100% Offline & Privacy-First: All transcription processing occurs on-device without an internet connection, ensuring voice data never leaves the computer.
Power Mode: Detects the active application or URL and automatically applies pre-configured settings for optimal dictation in that context.
Context Aware AI: Uses AI to understand the content visible on the screen and adapts the transcription behavior accordingly.
Global Shortcuts: Supports configurable keyboard shortcuts for starting/stopping recording and for push-to-talk functionality.
Personal Dictionary: Allows users to add custom words, industry-specific terminology, and smart text replacements to improve transcription accuracy.
Use Cases:
Developers, writers, or professionals needing to dictate text quickly into any macOS application without typing.
Privacy-conscious users who require a voice-to-text solution that processes speech entirely offline on a local machine.
Users who frequently switch between different applications (e.g., email, code editor, browser) and want automatically applied dictation settings for each context.
Why It Matters:
As an open-source project, VoiceInk provides transparency into its local AI operation and gives developers the ability to build and modify the application from source code. It is a direct alternative to cloud-dependent dictation services by offering 100% offline processing as a core feature, with a stated goal of being the most efficient and privacy-focused solution for macOS. The project is built on top of well-known open-source AI inference libraries like whisper.cpp.
