Voz
Voz is a free, open-source dictation app for macOS under MIT, by Joël Heil Escobar. Hold a shortcut, speak, release, and Voz types the text in the app under the cursor. NVIDIA's Parakeet model does the transcription, Apple's on-device model cleans it up, and audio and history stay on the Mac unless an extension sends them somewhere. One caveat: it needs macOS 26 on Apple Silicon.
What does Voz do?
Voz describes itself as "Type at the speed you talk." Hold a shortcut, speak, release, and the text is typed into whatever app has focus. The README says audio and transcripts stay on the Mac unless an extension sends them elsewhere, and that no account is required.
Two models do the work. Parakeet, by NVIDIA, transcribes 25 languages. Apple's on-device model, part of macOS 26, cleans the text up without a network round trip. Extensions can run on the dictated text, and extensions that use an AI model let the user pick between ChatGPT, Claude or Apple's model. Password fields and private browsing windows are handled separately so sensitive text is not captured.
Key facts
- Licence: MIT, copyright Joël Heil Escobar, 2026.
- Platform: macOS 26 on Apple Silicon only, from the README's "Requirements" section.
- Speech model: NVIDIA Parakeet, 25 languages.
- Cleanup: Apple's on-device model on macOS 26.
- Extensions: TypeScript, with a CLI and a Model Context Protocol server for coding agents to drive them.
- Audio and history: kept on the Mac by default. The README says audio and transcripts stay on the device unless an extension sends them elsewhere.
- Telemetry: anonymous counts with a random id; crash reporting on by default with a privacy setting to turn it off.
- Install: download the disk image from the Releases page, drag Voz to Applications, open and follow the setup.
What does it replace, and where does it fall short?
Voz replaces Wispr Flow for Mac users on macOS 26 who want a push-to-talk key, local transcription and no subscription.
Where it falls short: macOS 26 on Apple Silicon only. Intel Macs, older macOS versions, Windows and Linux are not supported. The cleanup model is Apple's on-device model, which is tied to macOS 26 and not available on earlier systems. Extensions that use ChatGPT or Claude for AI features send text to those providers. The README states that no word limit is set but gives no measured latency or accuracy figures.
How does Voz run?
Download the disk image from the Releases page, drag Voz to the Applications folder, open it and work through the setup wizard. Voz asks for Accessibility permission so it can type into other apps, and Microphone permission so it can record.
Parakeet and Apple's on-device model are downloaded on first run. After that, the push-to-talk shortcut works offline. Extensions that need an outside model prompt for which provider to use, and the key goes in the settings.
Who is Voz for?
A Mac user on the current macOS who wants a dictation key on every app, and who writes TypeScript or runs a coding agent and wants to automate what happens to the dictated text. Someone on Intel, on an older macOS, or on Windows and Linux should look at Handy, OpenDictate or Plume instead.
What limits does the README state?
- macOS 26 on Apple Silicon is required; older macOS and Intel Macs are not supported.
- Accessibility and Microphone permissions are required, with Input Monitoring for global shortcuts.
- Some extension capabilities are filtered for security: password fields are protected, private browsing windows are handled separately.
- Extensions that use ChatGPT or Claude send text to those providers; Apple's on-device model is the only choice that keeps everything on the Mac.
- Telemetry and crash reports are on by default and can be turned off in Privacy settings.
Questions people ask
Is Voz open source?
Yes. The LICENSE file at the repository root is MIT, an OSI-approved licence, copyright 2026 Joël Heil Escobar. The README calls the project free and open source and says there is no account to create. The downloaded models have their own licences: Parakeet is NVIDIA's open-weights release, and Apple's on-device model is part of macOS 26.
Does Voz work offline?
Yes, for the default path. After the first-run model download, dictation works with no internet, and the README says audio and transcripts stay on the device. The exception is an extension that calls ChatGPT or Claude, which sends text to that provider; those extensions use a key the user provides, and Apple's on-device model is the choice that keeps everything on the Mac.
Can Voz be used commercially?
Yes. MIT allows commercial use of the code without a fee, as long as the licence text travels with copies. The speech and cleanup models carry their own licences, so a commercial user should also read the licences for Parakeet and the Apple on-device model. The README does not add any extra restriction beyond the MIT grant.
How does Voz compare with Wispr Flow?
Voz does the core of Wispr Flow, speak and paste, with no subscription and with audio kept on the Mac. Wispr Flow runs on older macOS and on Windows; Voz needs macOS 26 on Apple Silicon. Wispr Flow is one product and Voz adds TypeScript extensions and an MCP server for coding agents. The open-source alternatives to Wispr Flow page sets it beside Handy, VoiceInk and others.
What makes Voz different from other open dictation apps?
Two things, both from the README. The cleanup step uses Apple's on-device model, so the text is tidied without a network round trip and without a second model download. And extensions are TypeScript modules with an MCP server, so a coding agent such as Claude Code can drive them. Other open dictation apps focus on transcription alone, with cleanup as an extra external step.
Sources
- Voz README and LICENSE: github.com/Joelheile/voz, read 2026-11-09.
More voice dictation apps
A free, offline push-to-talk dictation app for macOS, Windows and Linux that runs Whisper or NVIDIA Parakeet on your own machine and pastes the text at your cursor, under MIT.
A GPL-3.0 macOS dictation app that runs Nemotron, Parakeet, Cohere Transcribe, Apple Speech or Whisper on the device, with command and write modes, per-app prompts, and an optional closed local enhancement model.
A dictation and meeting-notes app under MIT that runs Whisper or Parakeet locally or cloud models with your own key, detects Zoom, Teams and FaceTime calls, and labels speakers on device.
A macOS dictation app under GPL-3.0 whose README names Superwhisper and Wispr Flow as the tools it replaces, with local models, per-app modes, a personal dictionary and an AI assistant mode.
A speech-to-text app inside the Epicenter monorepo under AGPL-3.0 that records, transcribes with a provider you choose, cloud or local, optionally polishes the text, and pastes it at the cursor.
An Apache-2.0 macOS dictation app that calls Gemini 3.5 Transcribe with your own key, writes audio to disk from the first millisecond for crash safety, and inserts polished text at your cursor without an account.
Added November 9, 2026. Every claim above comes from the project's README, LICENSE or model card, read on November 9, 2026, or from the GitHub API on the date shown in the panel. Found an error? Write to reveneau@licheo.com and it is fixed in the next weekly pass. Repository: github.com/Joelheile/voz.
Want Voz running on your own servers?
Reveneau sets it up on your hardware, connects it to your product, and builds the software around it.