Two private dictation apps with different ideas about polish.
Both can keep speech and cleanup on your device. Both are free and open source for personal or work use. Vox adds editable Voice Modes, Windows support, a "Hey Vox" wake word, and a voice-controlled task list.
Quick answer
Vox or EnviousWispr: which should you choose?
Both Vox and EnviousWispr can transcribe speech and clean up text locally on Apple Silicon Macs. Choose Vox if you want editable Voice Modes, automatic app-based mode suggestions, Windows support, or the VoxNotch voice-controlled task list. Choose EnviousWispr if you want direct paste, Escape Recovery, custom-word correction that works with either engine, and the option to use local models or your own cloud provider. Vox is the more mode-driven product; EnviousWispr is one path from speech to pasted text. Both are free for personal and work use. Choose based on the workflow you will actually use: Vox rewards people who create multiple modes and use tasks, while EnviousWispr keeps dictation and polish in one path.
Still deciding? See the best Mac dictation apps by use case or browse all 17 dictation app comparisons.
Feature comparison
The short version: both keep speech recognition local. Cleanup privacy, licensing, modes, platforms, and how the finished text reaches your app are where the choice changes.
| Feature | EnviousWispr | Vox |
|---|---|---|
| Price and license | Free for personal and work use; open source under GPLv3 | Free for personal and work use; open source under MIT |
| Platforms | Apple Silicon Mac, macOS 14+, and Android 13+ | Mac and Windows; Apple Silicon Mac with macOS 14+, or Windows 10/11 |
| Speech engines | Parakeet Ultra and Whisper | Parakeet and Whisper, with several local model choices |
| Audio sent to a server | Never | Never |
| AI cleanup | EG-1, S1-mini by Superwhisper, Apple Intelligence, or a downloaded Ollama model | Downloaded Gemma 4 or Apple Intelligence |
| Cloud cleanup | Optional OpenAI, Gemini, Claude, or hosted Ollama with your own account or key | No third-party cloud provider option |
| Writing modes | One cleanup pipeline that preserves the dictated tone | Editable Voice Modes for chat, email, notes, code comments, summaries, and custom styles |
| App-aware behavior | App context helps polish the transcript | Suggested and automatic modes based on the current app |
| Finished text | Can paste directly or copy to the clipboard | Copies to the clipboard by default; the installed settings also expose automatic paste |
| Voice task list | No | VoxNotch can add, finish, reorder, reprioritize, and remove local tasks by voice on Mac |
| Custom vocabulary | Yes, with aliases and Smart Import from other apps | Yes, a local personal dictionary (not used by its Parakeet engine) |
Vox claims were checked in Vox 1.10.0 and on Vox's official site. Pricing was checked on its official Teams page. Competitor claims last verified: 2026-10-17.
What both apps get right
Choosing local dictation no longer means giving up cleaned-up text.
Both transcribe on your device. Neither needs to upload your recording for ordinary dictation.
Both can remove fillers, fix punctuation, and shape the result with an on-device writing model.
After model downloads finish, either app can handle the full dictation path offline when a local cleanup model is selected.
The clearest reasons to choose each one
EnviousWispr can paste directly where you were writing, recover an accidental cancellation, and import custom words from other dictation apps. It corrects your custom words in whatever the speech engine writes, Parakeet included; Vox's own notes say its Parakeet engine does not use Vox's custom dictionary.
You can keep polish local or bring your own cloud provider. That gives you a simple default without closing off model choice later.
Vox gives each writing context its own editable prompt. It can suggest a mode for the current app, and its built-in choices cover chat, email, notes, code comments, summaries, and more.
VoxNotch is a separate local task surface controlled by natural speech, and an opt-in "Hey Vox" wake word starts dictation hands-free. Vox also runs on Windows, which EnviousWispr does not.
Frequently asked questions
Yes. Both apps are now free for personal and work use, and neither requires an account. EnviousWispr is open source under GPLv3; Vox is open source under the MIT license. Vox accepts optional contributions.
Yes, when using downloaded local models. Vox can transcribe locally and use downloaded Gemma 4 for local cleanup. Vox also offers Apple Intelligence on supported Macs. EnviousWispr can use local EG-1, S1-mini by Superwhisper, Apple Intelligence, or downloaded Ollama models, plus optional cloud providers with your own key.
Yes. Vox can clean up text locally with a downloaded Gemma 4 model. It also offers Apple Intelligence on supported Macs. Vox's built-in and custom Voice Modes control how cleanup behaves. EnviousWispr offers EG-1, S1-mini by Superwhisper, Apple Intelligence, or a downloaded Ollama model locally, plus optional cloud providers with your own key.
It is a local task list that sits at the top of the Mac screen. You can add, finish, remove, reorder, and reprioritize tasks by speaking naturally.
Yes. Vox is now open source under the MIT license, with its source on GitHub. EnviousWispr is open source under GPLv3.
Both support Apple Silicon Macs starting with M1 on macOS 14 or later. Neither supports Intel Macs. Vox also supports 64-bit Windows 10 and 11, while EnviousWispr has an Android app instead.
Compare more on-device dictation apps:
Want direct paste and Escape Recovery?
No account. No subscription. Your audio stays on your Mac.