Inbox
The 40-email morning
A small-business owner answers 40 emails of about 80 words each. Hold, answer in one breath, let go, next.
- Typed
- 62 min
- Talked
- 25 min
Talk to me, I'll type.
Hold two keys, talk, let go. Your words appear in any app, about three times faster than typing. Talk2me turns speech into text on your Mac's own chip, so your voice goes to no server: not ours, not anyone's.
Free in beta, then $10 a year. No card. Mac with Apple silicon.
Have an invite code? Enter it hereFollowing our call, the client agrees to settle at $240,000, provided the confidentiality clause stays as drafted.
Why privacy first
Your voice is the most personal data you type with. It carries your words, the sound of you, and the names of your clients, patients and sources. Most AI dictation apps send it to a server to turn it into text. Talk2me does that work on your Mac instead.
Where your voice goes
Cloud apps differ: some let you turn training off, some keep no copies. Ask each one the four questions on the left. How to check any dictation app
Proof, not promises
Then dictate. It still works, because nothing needs a server. (After the one-time voice model download.)
There is none. No account, no user ID, no password to leak. Nothing ties your words to you.
One switch in Settings keeps every step on your Mac, including the optional text cleanup.
Prefer a cloud engine? Every route in Settings is tagged 100% private, Private · ZDR or Not private, before you choose it.
Our privacy policy lists everything we keep, in one table. No voice, no words, no IP addresses.
No tracking cookies, no analytics company, no third-party script or font. Your browser's developer tools will show you.
How it works
No new app to learn. One shortcut, in every app you already use.
Hold two keys. From any app.
Talk like you talk. Pause, think, carry on. Up to 10 minutes at a time.
Let go. Your words land at your cursor, already punctuated.
Fast
The average person types about 52 words a minute. Most people talk two to three times faster than that. Talk2me turns the difference into time you get back. Try your own numbers:
Your estimate
6 h back each week
You type about 6,240 words a day. Saying them takes 48 minutes instead of 2 hours.
About 288 hours a year, over 48 work weeks.
An estimate. Typing speed: average of 168,960 typists, Dhakal and others, CHI 2018. Stanford researchers found speech about 3 times faster than typing on a phone. More on voice typing productivity
Email, Slack, documents, your code editor, AI chats. If you can type there, you can talk there.
"Twenty five" becomes 25. Dates, amounts and times come out the way you would type them.
A personal dictionary spells client names, drug names and jargon your way.
Optional cleanup removes the "ums" and fixes punctuation, on your Mac if you prefer.
Whisper Large v3 Turbo on Macs with 16 GB of memory, sized down below that. Cohere Transcribe is one click away.
Every update is signed and checked. Nothing installs until you say so.
Stories
Three everyday examples of how people use voice typing, with the time each one takes.
Inbox
A small-business owner answers 40 emails of about 80 words each. Hold, answer in one breath, let go, next.
Legal
A lawyer dictates a 400-word file note right after a client call, while the details are fresh. The note stays on the Mac.
Clinic
A clinician writes up 10 visits of about 150 words each, in the record system, before going home.
Examples of use, not customer quotes. Times use 52 words a minute typed and 130 talked; yours will differ. In the beta? Tell us how you use Talk2me. See all 5 examples
Who it's for
Dictate memos and client emails without sending privileged words to a third party.
Dictation for lawyersWrite notes at the speed of speech, with patient details that stay on your Mac.
Dictation for healthcareDraft and take notes without leaving a copy of your sources on someone else's server.
Dictation for journalistsTalk your prompts, commit messages and docs. Your code context stays local.
Dictation for developersCompare
| # | Question | Talk2me | Wispr Flow |
|---|---|---|---|
| 1 | Where your voice becomes text | Good: On your Macby default | Concern: Their cloudalways, per their docs |
| 2 | Your words used to improve AI | Good: Never | Concern: On by defaultyou can turn it off |
| 3 | Reads your screen | Good: Text before the cursorstays on your Mac | Concern: On-screen texton by default |
| 4 | Works offline | Good: Yes | Neutral: Noneeds internet |
| 5 | Account needed | Good: No | Neutral: Yes |
| 6 | Price per year | Good: $10free in beta | Neutral: $144$180 paid monthly |
From Wispr Flow's own pages, checked 30 September 2026. Superwhisper too
Price
Private beta
Free
After the beta
$10 a year
Why so little? Your Mac does the heavy work, so we pay for no servers to process your voice. We would rather charge a small fair price than sell your data.
FAQ
Yes. With a voice model on your Mac (the default), your audio becomes text on your computer and never leaves it. Talk2me has no account and collects nothing about you. If you choose a cloud engine instead, your audio goes straight from your Mac to that provider, with your own key, and Settings tags the route "100% private", "Private · ZDR" or "Not private". One switch, Strict privacy, keeps every step on your Mac.
Never. We never receive your voice or your words, so we can't train on them, and we never will. Not with your consent, not anonymized, not later.
Only to download the app and, once, the voice model. After that, dictation works with no internet at all. Talk2me checks once a day for a signed update file; that request carries nothing about you.
A Mac with Apple silicon (M1 or later) and macOS 15 or later. The app is about 150 MB to download, plus the voice model picked for your Mac (about 0.5 to 1.6 GB, once). Windows and Linux are next: join the list and we'll email you.
Talk2me runs state-of-the-art speech models on your Mac. It starts on Whisper, sized to your Mac: Large v3 Turbo with 16 GB of memory or more, Small below that. A small starter voice lets you dictate while it downloads. With 16 GB or more, Settings > Voice also offers Cohere Transcribe on the GPU. A personal dictionary spells your names and jargon your way.
We are a small team in Montréal. The beta is small so we can listen closely and fix things fast. Join the list and we'll send your invite as spots open. Your email is used for that and nothing else.
By Wispr Flow's own documentation: your voice is always transcribed in the cloud, model training on your dictation is on by default for Free and Pro, and Context Awareness reads on-screen text by default. Talk2me keeps your voice on your Mac, never trains on it, and costs $10 a year. See the full comparison, with sources.
We keep it in our own database, only to send your invite, and never share or sell it. Remove it yourself, any time:
Join the free beta. We'll email your invite when a spot opens. No newsletter, no sharing.
Have an invite code? Enter it here