VoiceInk Review

VoiceInk runs transcription on your Mac using local models, so audio stays on the device and core use works offline. Its README acknowledges whisper.cpp, Parakeet through FluidAudio, local GGUF models, and SenseVoice Small. On top of transcription it offers Modes, which switch settings and AI prompts by active app or website, a personal dictionary with text replacements, optional screen or clipboard context, configurable shortcuts including push-to-talk, and a voice assistant mode. Optional cloud AI enhancement receives text only, not audio.

You can install it with Homebrew (brew install --cask voiceink), download the compiled app from the website with a free trial, or build it from source using the repository's BUILDING.md. The README states macOS 15.0 or later, while the OpenAlternative listing mentions macOS 14.4 or later and Apple Silicon, so check the website for the current requirement before buying. The maintainer is not accepting pull requests, though bug reports and forks are welcome.

The workflow is hold a shortcut, speak, and the text appears in whichever app has focus, with Modes changing tone or cleanup per app. It is a native Mac app rather than a web service, so there is no account to manage for local use. We did not run it ourselves; this section reflects the repository README and OpenAlternative's description.

• On-device transcription with several local model options
• Modes that auto-switch by app or website
• Personal dictionary and text replacements
• Global shortcuts with push-to-talk
• Screen-aware context and an AI assistant mode

• Privacy by design, since transcription stays on the Mac
• One-time payment instead of a subscription, with source available to build yourself
• Active development, with a release in late September 2026

• Mac only and tied to recent macOS and Apple Silicon
• The convenient compiled version with updates and support is paid, although the source is free
• The project does not accept pull requests, and the README does not publish a language list
• Accuracy claims such as 99% come from the project and depend on the model you choose

Building from source is free under GPL-3.0. The compiled app on the project's website is a one-time purchase with lifetime updates and a free trial, and the README does not state the trial length or the price, so check the site. There is no subscription.

Wispr Flow is a cloud-based dictation service with a subscription, while VoiceInk does the speech recognition on your own Mac and charges once. Superwhisper is the closest local competitor with a similar on-device approach. VoiceInk's extra argument is that its source is public and buildable, which matters if you care about auditing what hears your voice.

• Mac users who want private, offline dictation without a subscription
• People who dictate in many apps and want per-app tone and cleanup
• Anyone who wants to inspect or build the app themselves

• Windows, Linux, or Intel Mac users
• Teams that need a managed, cross-platform dictation service
• Anyone who wants the app free without compiling it