# VoiceInk review: a GPL-licensed macOS dictation app you build from source > VoiceInk is a native macOS voice-to-text app built on whisper.cpp and local GGUF models. The download is paid, the source is GPL v3, and pull requests are closed, which shapes who can realistically use it. **Beingpax/VoiceInk** — The best open-source alternative to Superwhisper & Wispr Flow. Voice-to-text app for macOS with no subscription - Repository: https://github.com/Beingpax/VoiceInk - Website: https://tryvoiceink.com - Stars: 6,617 · Forks: 936 - Language: Swift - License: NOASSERTION - Published: 2026-09-22 · Updated: 2026-09-22 · Language: en - Canonical page: https://hysenlabs.com/projects/beingpax-voiceink ## The problem VoiceInk solves, and who it is actually for Dictation on macOS has two common shapes. Cloud services give good accuracy but send audio off the machine. Local tools keep audio on disk but often stop at a transcript file. VoiceInk sits in the second group and adds the parts that make dictation usable inside other apps: a global shortcut to start and stop recording, modes that switch settings based on the frontmost app or URL, a personal dictionary, and text insertion into whatever field has focus. The README describes the goal as "the most efficient and privacy-focused voice-to-text solution for macOS" and claims local models transcribe "with 99% accuracy, almost instantly". Treat that figure as a vendor claim, not a measured result. The audience is narrower than the marketing suggests. The README states the project is not accepting pull requests, so anyone who wants to fix a bug in the app itself has no upstream path, only a fork. That single line matters more than any feature list when you are deciding whether a team can depend on it. ## How the transcription pipeline is put together VoiceInk is a Swift macOS app, and the repository layout shows a single Xcode project (VoiceInk.xcodeproj) plus a separate VoiceInkRefineXPC target. The XPC target is the interesting part: an app that inserts text into other applications and holds a global shortcut is a good candidate for splitting privileged or long-running work into an XPC service, and the target name suggests exactly that for the refine step. Inference does not live in the repository. The acknowledgments list whisper.cpp for Whisper inference, FluidAudio for the Parakeet model implementation, and a SwiftPM distribution of transcribe.cpp for local GGUF models, with SenseVoice Small from FunAudioLLM as a multilingual option. So the app is a shell around several interchangeable local engines rather than one bundled model. Supporting dependencies include Sparkle for updates, KeyboardShortcuts for user-configurable bindings, LaunchAtLogin, MediaRemoteAdapter for pausing media during recording, and SelectedTextKit for reading selected text. The Makefile confirms the build-time dependency: it clones ggerganov/whisper.cpp into $HOME/VoiceInk-Dependencies and runs build-xcframework.sh to produce whisper.xcframework, which the Xcode project then references. ## Installing VoiceInk on macOS: Homebrew cask or your own build The README points at tryvoiceink.com for the compiled app, described as a free trial, with a purchase that the author says funds continued development and unlocks automatic updates and priority support. If you want the compiled route without leaving the terminal, the README gives a Homebrew cask. The cask installs the signed build, so Sparkle can handle updates afterwards. ```bash brew install --cask voiceink ``` The source route is different and worth reading before you start. The README defers to BUILDING.md, and the Makefile shows what that involves: it checks for git, xcodebuild and swift, clones whisper.cpp into $HOME/VoiceInk-Dependencies, and builds the framework there before touching the Xcode project. The default target runs the check and then the build. ```bash make all ``` For a development loop the Makefile defines a dev target that builds under the scheme name VoiceInk Dev and then launches it, and a local target that signs with a local Apple Development identity when one is available. Expect the first build to be slow, because the whisper framework is compiled from source rather than downloaded. The repository also ships a LocalBuild.xcconfig, which is where local build settings belong rather than in the shared project file. ## Where VoiceInk breaks down or is the wrong choice Platform is the first wall. The README requires macOS 15.0 or later, and the topics list macos, macos-app and swift. There is no Windows, Android or iOS target in the repository layout, and nothing in the README promises one. If you dictate on a phone, VoiceInk is not the tool. The second issue is contribution. The README is explicit that the project is not accepting pull requests and suggests forking instead, with bug reports and documentation issues as the accepted channels. A team that needs a fix landed upstream is choosing a fork it must maintain. Third, the licence situation is inconsistent on the surface. The README links a GPL v3 badge and states GNU General Public License v3.0, while the repository metadata reports NOASSERTION. GPL v3 is a copyleft licence, so if you plan to ship VoiceInk inside a proprietary product, the terms matter and the discrepancy is something to resolve by reading the LICENSE file rather than the badge. Finally, the accuracy claim in the README is unverified here; local model quality depends on which engine and model you select, and the repository does not publish a benchmark. ## VoiceInk compared with MacWhisper and other dictation tools The comparison people search for is VoiceInk against MacWhisper, and the difference is mostly packaging and licence rather than the underlying model. Both lean on Whisper-family local inference, so raw transcription quality is unlikely to be the deciding factor. MacWhisper is a closed commercial application; VoiceInk publishes its Swift source under GPL v3 and lets you build it, while charging for the compiled build with updates and support. That means with VoiceInk you can audit and modify the app, and with MacWhisper you cannot, but with MacWhisper you get a product with no build step and no Xcode requirement. Against cloud dictation services, the split is different again: VoiceInk's README states processing is 100% offline, so audio does not leave the machine, at the cost of managing local models and disk space yourself. The repository also points at FluidAudio, transcribe.cpp and SenseVoice Small, which suggests engine choice is a first-class part of the app rather than a fixed pipeline, something a single-model tool does not offer. ## Maintenance, releases and what upgrading costs you The last push to the default branch was on 2026-09-21, and the most recent releases listed are v2.20 on 2026-09-19, v2.13 on 2026-08-27 and v2.11 on 2026-08-12. The version numbers move quickly between releases, which is consistent with frequent small updates rather than long release cycles. The repository is not archived. For a compiled install, the README ties automatic updates to the purchased build, and the dependency list includes Sparkle and an appcast.xml at the repository root, so update delivery is part of the app rather than something you script. For a source build, upgrades are your responsibility: you pull the repository, and the Makefile's whisper target will reuse an existing whisper.xcframework if one is already present in $HOME/VoiceInk-Dependencies, updating the whisper.cpp checkout only when the framework is missing. That means a stale framework can survive an app upgrade unless you remove it. Keep that in mind before filing a bug against a source build. ## Frequently asked questions about VoiceInk The questions below come from search data and from what a new user of this project would reasonably ask. Answers stay within what the README, the Makefile and the repository layout state. Where the material is silent, the answer says so rather than guessing. ## Conclusion Adopt VoiceInk if you are on macOS 15.0 or later, want transcription to stay on the machine, and either accept the paid compiled build or are prepared to build the Xcode project yourself. Do not adopt it if you need Windows, Android or iOS, if you need to contribute patches upstream, or if you want a permissive licence you can fold into a closed product. Before committing, open BUILDING.md and the Makefile, confirm your Xcode can build the whisper.xcframework, and read the LICENSE file for the exact terms, since the repository metadata reports the licence as NOASSERTION while the README states GPL v3.0. ## FAQ ### Is VoiceInk free? The source code is available under GPL v3 and you can build it yourself, but the compiled app from tryvoiceink.com is described as a free trial with a purchase that the author says funds development and adds automatic updates and priority support. ### Is VoiceInk worth it? That depends on whether you value local processing and a buildable codebase over a finished product. The README claims 99% accuracy with near-instant local transcription, but the repository publishes no benchmark, and the compiled build with updates and support is paid. ### How do I use VoiceInk? Install the compiled app from tryvoiceink.com or via brew install --cask voiceink, then trigger recording with a global shortcut and let VoiceInk insert the transcription into the focused field. Modes and the personal dictionary change how the text comes out per app. ### Does VoiceInk work on iPhone? No. The README lists macOS 15.0 or later as the requirement, and the repository is a Swift Xcode project with no iOS target. ### What is VoiceInk? It is a native macOS voice-to-text application that transcribes speech locally, using engines such as whisper.cpp, FluidAudio for Parakeet and transcribe.cpp for local GGUF models, with global shortcuts and app-aware modes. ## Sources - [Beingpax/VoiceInk on GitHub](https://github.com/Beingpax/VoiceInk) - [Issues](https://github.com/Beingpax/VoiceInk/issues) - [Project website](https://tryvoiceink.com) - [README](https://github.com/Beingpax/VoiceInk/blob/main/README.md) - [Releases](https://github.com/Beingpax/VoiceInk/releases) --- Hysen Labs editorial analysis, written from the project's own repository and release notes. Cite the canonical page: https://hysenlabs.com/projects/beingpax-voiceink