Yes, VoiceInk is open source: its code is on GitHub under the GNU General Public License v3.0, and you can build the Mac app yourself. The ready-made app sold on tryvoiceink.com is the same project under a separate commercial licence, with automatic updates and support. On privacy, VoiceInk's policy says "By default, all transcription processing happens entirely on your device using local AI models", and that cloud transcription and AI enhancement only run if you add a provider's API key and select them. VoiceInk lists no security audits, so "safe" here rests on its documented local design and on the settings you choose.
Publisher note: SuprFlow publishes this guide and is one of the products discussed. Competitor details come from each company's own pages, checked on 1 October 2026. Prices and features change; confirm before you buy.
This page reports what VoiceInk's own pages and repository say. It is not a security audit or a code review, and we have not tested VoiceInk's network traffic.
What do VoiceInk's own pages say?
As of 1 October 2026.
| Question | What VoiceInk states | Source |
|---|---|---|
| Who makes it? | A developer who signs the project as Pax, under the GitHub account Beingpax. | GitHub README |
| Where is speech turned into text? | On your Mac by default, with local models such as Whisper and Parakeet. Cloud transcription is optional and sends "your audio file" to the provider you select. | Privacy policy |
| What about AI enhancement? | Optional. "Only the transcribed text (never audio) is sent to your selected provider", plus vocabulary and any context you enable. VoiceInk Refine cleans text locally. Dictionary Auto Learn sends changed text and nearby text to its own selected provider. | Privacy policy, Privacy and data docs |
| What is stored, and where? | Transcripts, enhanced text and saved audio stay on the Mac, kept "indefinitely by default", with optional automatic deletion. API keys are in the macOS Keychain. "We do not store your app transcriptions or audio files on VoiceInk servers." | Privacy policy |
| Training use | Not addressed directly. With cloud providers, "data retention at the provider's end is governed by their individual privacy policies". | Privacy policy |
| Audits and certifications | None listed on the pages we read. | — |
| Platforms | Mac, plus a separate iOS app. The docs and README list macOS 15 or later; the pricing page lists Apple Silicon and macOS 14.4 or later. | Installation docs, Pricing |
| Source availability | GPL v3.0. The README says the project is "not accepting pull requests at this time". | GitHub README and LICENSE |
Supported cloud providers listed in the privacy policy include Groq, OpenAI, Anthropic, Gemini, Mistral, Cerebras, Deepgram, ElevenLabs, Speechmatics, Soniox and OpenRouter, plus Ollama, which runs locally, and custom OpenAI-compatible endpoints.
Paid app or free source build: what is the difference?
VoiceInk's terms describe "a dual-licensing model". Under GPL v3 you may access, modify and "Compile and run the application from source". The pre-compiled app from the website is under a commercial licence covering a set number of devices, with automatic updates and support. VoiceInk pricing lists the tiers.
What changes if you build it yourself, according to BUILDING.md:
- You need macOS 15.0 or later and Xcode with Command Line Tools.
- "Local builds do not include iCloud dictionary sync or automatic updates."
- Ad-hoc signed builds "may require macOS permissions again after rebuilding".
One caution applies to every open-source app: public source lets you read the code, but it does not prove that a downloaded binary was built from that code. Building it yourself is the way to be sure. What source and a licence can prove explains the limits.
What does VoiceInk send over the network?
With local models and no enhancement, the privacy policy says no transcription data leaves the Mac. Other traffic its pages mention:
- Licence activation. The licence docs say activation "checks your license with the licensing service and requires a working network connection".
- Updates. The README lists Sparkle, an update framework, for the compiled app.
- iCloud dictionary sync in the compiled app, per BUILDING.md.
- Model downloads before local transcription can run.
- Context. The privacy and data docs say VoiceInk prepares a local snapshot of clipboard, selected text and, with permission, on-screen text extracted by OCR. That context is sent only in enhancement requests whose mode includes it.
What do VoiceInk's pages not say?
- Whether the compiled app sends usage analytics. The privacy policy covers transcription data, storage and cloud providers; it does not mention app analytics either way.
- Audit or security review results. None are listed.
- One wording difference. The README's feature list says "100% offline processing ensures your data never leaves your device". The privacy policy and docs are more precise: local by default, with opt-in cloud features. Read the policy as the fuller description.
How can you check for yourself?
- Download from tryvoiceink.com, as the installation docs say, or build from the GitHub repository. The README also lists a Homebrew cask. How to check a signed Mac release shows how to see who signed the app.
- Review permissions. The docs list Microphone, Accessibility (pasting, selected-text context and Auto Learn) and, optionally, Screen Recording for screen context.
- Check all three settings the docs name: transcription model, enhancement provider and Auto Learn's review provider. For an on-device setup, the docs advise local transcription, no enhancement or a local option, and Auto Learn off or pointed at a local Ollama model.
- Set history and audio retention in History settings.
- Run a network test with fictional text. Turn off Wi-Fi and dictate "Ship order 3392 to Hana Kowalczyk, 2,710 dollars, Tuesday." If it transcribes, speech recognition is local.
When is VoiceInk a reasonable choice, and when does something else fit better?
VoiceInk is a reasonable choice when you want local dictation at a low one-time price, value being able to read or build the source, and are comfortable choosing your own models and providers. Its public source lets a technical user inspect or build exactly what runs.
Something else may fit better if you need published audit reports for a vendor review, an Intel Mac (the docs say local models do not work reliably there and recommend cloud), Windows, or a team admin console.
SuprFlow is one alternative, though not on transparency: it is proprietary, so you cannot read its source. It runs speech recognition on the Mac with a downloaded WhisperKit model, and its Default formatting is local cleanup of punctuation, capitalization and numbers; optional Cloud API formatting sends transcript text, not audio, to a provider you configure. It uses the network for licence checks, model downloads, updates and usage telemetry. It is Mac-only (Apple Silicon, macOS 14 or later), costs $125 once with no free tier or trial, and has no audit reports. VoiceInk costs less and has a trial. SuprFlow vs VoiceInk compares them.
Questions people ask
Is VoiceInk free?
The source is free under GPL v3 and you can build it yourself. The ready-made app has a 7-day trial, then a one-time licence. See VoiceInk pricing.
Is VoiceInk good?
We have not benchmarked it, so we will not rate it. It documents local Whisper and Parakeet models, optional cloud providers, modes and context awareness. Try the trial with your own voice.
Where is VoiceInk on GitHub?
The repository is github.com/Beingpax/VoiceInk, which tryvoiceink.com links to as its source code. Other GitHub projects use similar names, so check the account name before downloading or building.
Does VoiceInk work offline?
Yes, with local models, once they are downloaded. Cloud transcription, cloud enhancement and licence activation need a connection.
Related guides
- VoiceInk pricing
- VoiceInk alternatives
- SuprFlow vs VoiceInk
- Open-source Wispr Flow alternatives
- What source and a licence can prove
- On-device dictation vs no internet