The cursor sits in the body of an email. Press the recording shortcut, say a few words, press it again to stop, and the text lands where the cursor was. No switching windows, no copy and paste. VoiceInk has turned a direct path into a product: it turns what you say into text at the cursor. Its sales model is more worth noting than its features—the source code is public on GitHub under GPLv3, and the README says the source is there for experienced developers to build; the compiled version starts at $25 as a one-time purchase.
To understand it, first look at which step it takes over for you, and what work remains for a person after that step.
What One Mode Decides
According to the official docs[1], the first use requires installation, going through onboarding, and then downloading a local model or connecting a cloud provider in AI Models. Next, choose a model in Mode—Mode also determines the language, whether to layer on AI enhancement, and how the result is delivered. Click the target text field, press the shortcut to start and stop, and that is the flow.
Results can be delivered in three ways: Paste inserts the transcribed text at the cursor; Respond shows the AI answer in the recorder; Custom Command hands the result to a local shell command. Beyond live speech, you can also queue existing audio and video files for transcription; results saved in History and Quick History can be searched, replayed, copied, pasted, exported, and retried.
This positioning is worth making clear: VoiceInk is primarily about instant dictation, not meeting notes. For example, if you want to reply with a paragraph in a GitHub issue, or edit a sentence in a chat box, your hands are slower than your mouth, and it shortens that path to pressing the recording shortcut to start and pressing it again to stop. Meeting recording and speaker recognition sit closer to the recording layer; MacWhisper's website[2] describes itself as an all-in-one transcription app for Mac, with file transcription, meeting recording, dictation, speaker recognition, export, 100+ languages, and support for both local and cloud models. The two overlap, but they are not the same product.
The Source Code Is Free; the Installed Version Is Paid
Public material about the person making the product is scarce. GitHub's README[3] is written in the first person: after 5 months of development, he decided to open-source it for the greater good; the source is there for experienced developers to build and contribute, while buying a license supports continued development and gets automatic updates, priority support, and future features. The same README also says that purchases help the author work on the project full time. The only attribution is “Pax” and the repository name Beingpax.
What can be confirmed is that the README lays out a somewhat awkward policy: the project currently does not accept pull requests, but forks and modifications are welcome, and the way to contribute is to report bugs, suggest features, and improve documentation through issues. Set next to “open source for the greater good,” this looks a little odd. But putting the two things together, you can see that he made the code public while keeping “installed, updatable, and someone to handle problems” on the paid side.
The pricing page[4] (retrieved 2026-11-14) divides licenses by device count: Solo $25 for 1 macOS device, Personal $39 for 2, Extended $49 for 3, including lifetime updates, and the page says 14-day refunds. The app only supports Apple Silicon Macs; the pricing page says macOS 14.4 or later, while the README says 15.0 or later—the two do not match. The terms are stricter: refunds must be requested within 14 days of purchase, with a valid reason, and the app must be uninstalled from all devices; after approval, the refund goes back to the original payment method in 5–7 business days. The pricing page's “no questions asked” and the terms' “valid reason” are not the same sentence.
Image: source VoiceInk[4], retrieved 2026-11-14; original interface retained.
The official pricing page sells individual licenses counted by device. The deliverable is a precompiled app plus updates and support, not the model itself; users pay cloud model costs to the provider themselves, and the terms[5] state that cloud services require your own API key. As for whether this business makes money, the materials contain no sales, revenue, refund rate, or paying-user count, nor model and API costs, so profit is completely unclear; this article does not estimate it.
There is another point that marketing language can easily glide past. “Local transcription” is true when you choose a local model: the audio stays on the Mac. But the documentation itself says that using only a local model does not mean the whole workflow is offline—optional AI enhancement sends the transcribed text to a cloud provider (not the audio), while optional cloud transcription sends the audio to the transcription provider the user chooses. The terms (updated 2026-04-20) use the same framing. This is not deception; rather, it spells out the boundary. But it means “100% offline” is a statement that holds only depending on configuration.
Traffic Rules in the Affiliate Program
On customer acquisition, there is one fairly hard rule. The affiliate program[6] pays 30% commission on qualifying sales, with 30-day tracking, monthly settlement, and a $100 minimum payout; joining is free and there is no audience-size requirement. But the same page prohibits paid search and bidding on VoiceInk keywords, paid social ads, sponsorships, display ads, and buying traffic; sales from paid traffic are not accepted, and related commissions are withheld. In other words, at least in this affiliate program, it is willing to share money with people who bring a real audience and unwilling to share money with bought clicks. It is a one-time charge, and the affiliate rules exclude paid traffic; but which channels actual growth comes from, the materials contain no conversion data, so it cannot be verified.
Also, do not write it as an “anti-subscription hero.” A one-time purchase is not unique in this category: MacWhisper's direct download version[7] is a one-time lifetime individual license, can dictate into any text field, and can detect meetings; what uses weekly/monthly/yearly subscriptions is Whisper Transcription on the App Store, which also has fewer AI provider options. So the difference is not the price form of “buyout,” but that it puts open source and instant dictation in the same product.
So the $25 buys a license for a precompiled app, along with automatic updates and technical support. A one-time charge collects the money on day one, but updates and support must continue to be delivered afterward; what is really being priced in the $25 is someone's willingness to keep doing this work.