Questions about dictating with VoiceSnap Pro

Use your dictation shortcut and speak. When you finish, punctuated text appears in the field your cursor was in — an email, a message, a ticket, a commit message or a browser form. Optional dictation history keeps a text copy on that computer. Dedicated voice notes use their own shortcut and save to your account for use across devices.

From the download page: pick the macOS or Windows installer, then sign in with Google or with a one-time code we email you. There is no password to create. Every feature is on from the first dictation, with a one-time free allowance of 20 minutes of transcription, 50 AI actions and up to 20 notes, and no card.

Click into the field you want the words to land in, hold the shortcut, and talk normally. Release the key and the finished text is typed in at the cursor. That is the entire workflow: no record button, no transcript pane, nothing to copy and paste.

Yes. The shortcut is set in VoiceSnap Pro's settings and you can pick whatever combination doesn't collide with the apps you live in — a spare function key, a modifier combination, whatever your hands already know. If something else on your machine has already claimed it, the keyboard shortcuts page in the documentation walks through choosing a replacement.

Yes. Click into the field where you want the text before starting. If insertion fails after transcription, use the Paste last dictation shortcut or find the text in local dictation history if you have it turned on. Dictations are not automatically saved as account-synced notes.

No. There is no enrolment step and no sample paragraphs to read aloud before you can start. Install it, grant microphone access, pick a shortcut, and dictate. The only thing worth teaching VoiceSnap Pro is your custom vocabulary — the names, product names and acronyms no general model could be expected to guess.

The built-in tools are free and can be enough for a short sentence. VoiceSnap Pro adds personal vocabulary synced across your devices, a shortcut that pastes your last dictation again, a History pop-up of recent dictations, voice notes saved to your account with AI actions and reminders, and the same app and shortcut on macOS and Windows. The comparison pages linked in the footer cover the differences in more detail.

More than 50, covering the widely spoken European, Asian and Middle Eastern languages you would expect from a current speech model. You don't select one before you start — VoiceSnap Pro works out what it is hearing. The definitive list ships with the app and is published on the languages page in the documentation.

Automatic language detection can help with different languages, but recognition can vary when you mix them in one sentence. Review bilingual text before sending it, and choose a language in Settings if automatic detection needs help.

Yes. The transcript comes back punctuated, so you can speak naturally instead of reciting punctuation marks. Review the finished text and adjust it in the app where it lands if you need a particular layout.

VoiceSnap Pro types what the speech model heard, punctuated. It does not run a separate rewrite over every dictation, so a hesitation or a false start can come through. When you want a dictation tidied, run an AI action on it — Fix or Shorten, say, or one you wrote yourself — and the result is yours to keep or undo.

Yes — that is exactly what custom vocabulary is for. Add surnames, client names, product names, repo names, internal acronyms and any spelling a general model would get wrong, and VoiceSnap Pro uses your spelling instead of its best guess. Seed it once with a dozen entries — about five minutes of work — and add to it whenever you notice something come out wrong.

It should. VoiceSnap Pro uses a modern speech model trained on a wide range of speakers rather than the old per-user systems that had to be tuned to one voice, so there is nothing to calibrate. If one particular word is consistently wrong, it is almost always a name or a technical term rather than an accent problem — add it to your custom vocabulary and it stops happening.

On clear speech with a reasonable microphone, expect around 99% of words to come back correct, with proper nouns and jargon making up most of what doesn't. Accuracy falls in the predictable places: a noisy café, several people talking at once, or a laptop microphone two feet away with a fan running. A cheap wired headset in a quiet room beats an expensive microphone across the desk.

With a little help, yes. Framework names, CLI flags, library names and API terms are precisely the vocabulary a general model has never heard in context, so add the ones you use often. Dictation earns its keep on the English part of programming — commit messages, code comments, pull request descriptions, replies in review threads — rather than on typing punctuation-dense syntax character by character.

For prose, usually by a wide margin: most people speak at roughly 150 words a minute and type at a fraction of that, and the gap grows over a long message. The honest caveat is that speaking and composing at the same time takes a week or two to get comfortable with. The free speaking-versus-typing calculator under Free Tools puts real numbers on your own rates.

It is sent over an encrypted connection to our server, which passes it to OpenAI to be transcribed, and the text comes back. For a dictation, that is the end of its job: the audio is discarded. For a voice note, the recording is kept with the note in your account so you can play it back — for as long as your Keep recordings setting says (forever, 30 days, or not at all), and never after the note is deleted for good. Your audio is not used to train AI models, not sold, and not mined to target advertising at you. The privacy policy linked in the footer is the formal version, and the privacy and data page in the documentation explains it in plain language.

No — this is the honest answer rather than the comfortable one. Transcription runs on servers, so dictation needs an internet connection; on a plane with no Wi-Fi, VoiceSnap Pro will not produce text. On-device transcription is something we want to build, and until it exists we would rather say so plainly than let you find out at 30,000 feet.

No. Neither your audio nor the text it becomes is used to train models. That is a rule the product is built around, not a checkbox buried three levels into a settings pane that you have to remember to switch off. The one thing that is up to you: if you turn on AI assistants in Settings, the assistant you connect reads your notes, and what its provider does with them is set by that provider's terms, not ours.

No. The microphone opens when you hold the shortcut and closes the moment you release it, or, in hands-free mode, runs from one press of the shortcut to the next. There is no wake word, no always-on listening and no background recording — if you have not started a dictation or a note, nothing is being captured. Both macOS and Windows show their own microphone-in-use indicator while it is open, so you never have to take our word for it.

In the dictation history on your own computer, which you can search by wording rather than by date, turn off, or clear. It is never uploaded: the only time a dictation's text leaves your computer is when you run an AI action on it, which sends that text through our server to OpenAI and keeps none of it. It exists because the most annoying thing about dictating into a chat window is that the paragraph you spent two minutes composing scrolls away forever. Notes are different: they are saved to your account, with their recordings, so they appear on your other devices. The notes library page in the documentation covers exactly where each lives and how to remove it.

Only if you turn it on, and only on the computer where you do. Settings → AI assistants → Let AI assistants read my notes lets Claude Desktop, or another assistant that supports MCP, search and read the notes on that computer through a small read-only server built into VoiceSnap Pro. It shares text only — never recordings or anything in the Trash — and your dictation history only if you also switch that on. VoiceSnap's server is not involved; what the assistant reads is sent to the assistant's provider (Anthropic, for Claude Desktop) under that provider's terms.

AI tags, on by default, sends a new note's text and the names of your tags, so AI can pick up to three of the tags you made — it never creates one. AI actions, such as Rewrite, Summarize or Translate, send nothing until you choose one; then the text you run it on, from a note or a dictation, and your own action's prompt go through our server to OpenAI, and the result comes back to you. Our server keeps none of that text, only a count of the actions you run. OpenAI does not use any of it for training. Both switches are in Settings → AI and apply on every device signed in to your account.

That depends on what you dictate and who writes your compliance policy, so here are the facts to take to them: audio leaves your machine over an encrypted connection to our server in Ireland and is transcribed by OpenAI; notes and their recordings are stored, encrypted, in your account; unless you turn AI titles and AI tags off, a note's text is also sent to OpenAI to write its title and pick among your tags; an AI action sends the text you run it on, from a note or a dictation, only when you choose one; nothing is used to train models, and nothing is sold. If your rules say no audio may leave the device under any circumstances, VoiceSnap Pro is not the right tool for that material until offline transcription ships.

On macOS, no. Password boxes and credential prompts turn on Secure Input, and VoiceSnap Pro refuses to start a dictation while it is on — a protection worth keeping, not working around. On Windows, VoiceSnap Pro cannot yet tell a password field from any other field, so a password dictated into one is handled like any other dictation: the audio goes through our server to OpenAI to be transcribed, and the text is pasted and kept in your dictation history. Either way, a password spoken out loud is a bad idea on its own merits. Type those by hand.

Ask us and it goes. The delete-my-data page linked in the footer opens an email to the same inbox as support. Write from the email address you sign in with, and we delete your notes (the Trash included), their earlier versions, their recordings, your tags and your own AI actions, and the records of your devices by hand. Deleting a note yourself, in the app, moves it to the Trash; it leaves your account 30 days later, or at once with Delete forever or Empty Trash. There is no retention argument at the end of it.

Nothing to start. Every new account gets a one-time free allowance with every feature and no card. The Pro plan, which removes the limits, is $18 a month or $108 a year, which works out at $9 a month, 50% less than paying monthly. Prices are in USD; local taxes or VAT may be added at checkout depending on where you are.

Yes. Every new account gets a one-time free allowance: 20 minutes of transcription (dictation and voice notes combined, shared across your devices), 50 AI actions and up to 20 notes, on macOS and Windows. Automatic AI titles, AI tags and reminders set by voice don't count as AI actions. It does not reset, and there is no separate trial: the allowance is the trial.

No. The free allowance needs no card and has no deadline. You enter payment details only if you choose the Pro plan, and checkout is handled by Polar, our merchant of record.

Nothing you already have is taken away: your notes stay readable and editable, with or without a subscription, and you can export them as Markdown files. New work beyond the allowance needs the Pro plan: dictation and new voice notes once the minutes are used, AI actions you run once those are used, and notes past the first 20. When part of the allowance is used up, the app tells you and offers the Pro plan.

Unlimited transcription, unlimited AI actions and unlimited notes, on up to 3 devices signed in to your account, for $18 a month or $108 a year. Every feature is already in the free plan; Pro removes the limits. Pro's fair-use ceilings, such as a cap of 20,000 notes, are listed on the Notes page.

In the app. Sign in, open Settings → Account and upgrade when you are ready, choosing monthly or yearly billing. Checkout is run by Polar, our merchant of record, which charges your card, adds any VAT or sales tax that applies, and emails a receipt for every payment. This website never takes a payment.

Yes. There are no separate Mac and Windows editions: you sign in with the same account on each. The Pro plan covers up to 3 devices on one account, Mac and Windows in any mix; on the free plan, your allowance is shared across your devices.

Cancel the Pro plan any time in the customer portal, which Manage subscription in the app opens. It stays active until the end of the period you have paid for, and nothing renews after that. The same portal switches you between monthly and yearly billing or updates your card. Refunds follow the refund policy linked in the footer.

They stay yours. Your notes remain in your account, readable and editable on every device you sign in on, and Export Notes saves them as Markdown files whenever you like. Cancelling only means new work beyond the free allowance needs Pro again. Deleting your data is a separate request, covered under Privacy and data above.

Yes, on free and Pro alike. Bug fixes, accuracy improvements, additional languages and new features arrive as ordinary updates. There is no separate upgrade fee.

Not as a self-serve option yet: each person subscribes on their own account. If you need the Pro plan for more than a handful of people, email support with the number you're after and roughly when you need it, and we will sort something sensible out.

macOS, Windows — and nothing else at the moment. Both builds are developed together and get the same features rather than one trailing the other by six months. Minimum OS versions will be published with the release notes rather than guessed at now, and there is no Linux build planned today.

No, and there probably won't be. The whole trick — typing into whatever field already has focus, in any app — is something mobile operating systems deliberately do not let a third-party app do. Both mobile platforms already ship dictation inside the system keyboard, which is the right place for it there.

Yes. Anywhere a cursor blinks is a target: Terminal, iTerm, VS Code, Cursor, JetBrains IDEs, a text editor in a terminal. In practice you'll use it for the English parts — commit messages, comments, PR descriptions, replies to review comments — rather than for dictating brackets and semicolons.

Usually, with one caveat worth knowing. VoiceSnap Pro runs on your local machine and inserts text into the focused window, so a Parallels, VMware or Remote Desktop window normally receives it like any other keystroke. Some remote clients filter synthetic input for security reasons; if yours does, dictate into a local note and paste it across — and tell us which client it was, because that is a bug report we want.

Yes, and there is no extension to install. Gmail, Google Docs, Notion, Linear, Jira, ChatGPT, a WordPress editor, a plain comment box on a forum — VoiceSnap Pro puts text into the page the same way your keyboard does, so the page cannot tell the difference. The per-app pages under Dictate In cover the quirks of the popular ones.

Yes, and the reason is worth understanding. macOS blocks one app from putting text into another app unless you have explicitly granted Accessibility permission in System Settings under Privacy & Security — and typing into the app you are already using is the entire product. You will also grant Microphone access, and Input Monitoring so the shortcut is heard while another app has focus. All three are granted once and can be revoked in the same place at any time. On Windows there is no Accessibility equivalent: you grant microphone access at the first dictation and that is the whole of setup.

Whichever one is closest to your mouth. A wired headset, AirPods, a USB microphone or the built-in laptop microphone all work — the built-in one just picks up more of the room along with you. VoiceSnap Pro follows your system input device by default, and you can pin a specific device in settings so a headset connecting mid-session never changes the answer.

No. VoiceSnap Pro is dictation for one speaker — your microphone, during a dictation you start. It does not join calls, record meetings or separate speakers, and adding that would make it a different product. Meeting-recorder tools cover that job; the comparison pages in the footer explain where the line sits.

Check whether a text field was focused, the right microphone was selected, and — on macOS — Accessibility permission is still granted. If the transcript is in your optional local dictation history but the field stayed empty, transcription worked and insertion needs attention. Check permissions and the target field; the Paste last dictation shortcut can put the text into a field you select now.

Text follows focus. If you start dictating and then click into another window while you are still speaking, the words arrive wherever focus ended up — that is the same behaviour your keyboard has. Click first, then hold the shortcut, and the problem disappears.

Some apps grab key combinations before anything else on the system sees them: full-screen games, remote-desktop clients in capture mode, and a few IDEs with aggressive keymaps. Pick a different combination in VoiceSnap Pro's settings for the general case, and tell us which app swallowed the old one so we can document it.

By default it follows the system input device, and both macOS and Windows sometimes leave input on the built-in microphone when a headset connects mid-session. Set the device explicitly in VoiceSnap Pro's settings and it stops being a question. If you dictate from more than one place — desk and sofa — pin the device you use most and switch deliberately.

That is a vocabulary problem, not an accuracy problem, and it has a direct fix: add the exact spelling to your custom vocabulary once. A client's surname, an internal project codename, a product spelled with a capital in the middle — after you add it, the model uses your spelling instead of the plausible-sounding alternative. It is the highest-value five minutes of setup in the app.

This happens most often on very short utterances — two or three words simply isn't much evidence to go on, especially with words that exist in several languages. Speak a full sentence and it usually resolves itself. If you dictate a language you use rarely, check the language settings in the app; the languages page in the documentation covers the details.

Transcription happens over the network, so a slow or unstable connection shows up as a delay between releasing the key and seeing text. Check whether other things are slow at the same moment — a video call in the background is a common culprit. If VoiceSnap Pro is consistently slow on a connection that is otherwise fine, that is a bug worth reporting with the details listed on the support page.

Look in Dictation history on the computer where you spoke, if Keep dictation history was turned on. Search a phrase you remember. History keeps the newest 1,000 dictations and can be cleared, so older or deleted text may no longer be there. Dedicated voice notes are separate and sync through your account.

Email support and include everything on the bug-report checklist on the support page — your OS and version, the VoiceSnap Pro version, the app you were dictating into, and what you said versus what actually appeared are the ones people most often leave out. A complete checklist is usually the difference between a same-day fix and a week of back-and-forth questions.