Say it once.
It’s already written.
Dictation that sounds like you. Meeting notes that write themselves. Stay in the conversation.
No key. No setup. It works the moment you install it.
Free while in beta · macOS 14+ · Apple silicon (Intel without the on-device model) · notarized by Apple
The settle
Hold a key. Speak. Let go. The words land where your cursor was, cleaned up and in the right tone for the app you're in. If they can't land, they're on the clipboard.
Meetings, noted
One tap when the call starts. No bot joins. Press End and the decisions, the actions and the loose ends are already written down.
It shows its work
Every model call is listed, with what it cost when the provider reports it. Every number says what it assumes. Nothing here asks for your faith.
The arithmetic
A paragraph, twice.
Same forty-two words. One typed, one said.
Typed
0s
Said
0s
Speaking time is measured. Typing is assumed at 40 words a minute, and you can change it. The app always says which is which.
Meetings
Be in the room,
not in your notes.
One tap when the meeting starts. SayLess listens through your Mac, so nobody sees a stranger join. Press End and the note is already there. Every line points at the words that made it. Ask it something next week and the answer shows its sources, or says it has none.
- No bot on the call. There isn't one to see.
- It records when you say so. The audio goes once the words exist.
Decisions
- Ship the beta build to testers Friday; feature freeze Wednesday noon 1 source
- Onboarding copy moves to Priya; first draft up for review Tuesday 1 source
- Hold the pricing survey until the beta has been out two weeks 1 source
Privacy is the architecture
Nothing to trust.
Everything to verify.
Configure nothing and nothing is sent to us. Apple's recognition and, on Apple silicon, an on-device model do the work. Add a provider and the app names every route and every cost. Send a note to Slack and it goes to Slack. You are told first. The difference isn't that less leaves. It's that you can see what did.
YOUR VOICE · 106 DICTATIONS
The briefing voice
Short sentences, almost no fillers. Formal in mail, looser in chat. Learned from the dictations you never edited.
The difference
It doesn't sound like AI.
It sounds like you.
Generic dictation gives everyone the same voice. SayLess learns yours. Sentence length, formality, the words only you use. Nothing joins the profile until you approve it, and you can correct or delete any of it.
How it compares
Five things you can check,
not take on trust.
| What | SayLess | Bot notetakers | Cloud dictation |
|---|---|---|---|
| Something joins the call | Nothing. It listens through your Mac. | A bot joins as a participant. | Not a call feature. |
| An account to use it | None. Everything local works signed out. | Usually required. | Usually required. |
| Where the final text is made | On this Mac, on Apple silicon, unless you add a provider. | On their servers. | On their servers. |
| Notes cite the transcript | Every line, or it is not written. | Varies. | Not a notes feature. |
| An assistant reads your notes | On this Mac, read-only, only if you switch it on. | Varies. | Varies. |
The left column is checkable in the app: in the meeting screen, on the source chip beside every line of a note, and in Settings. The other two describe categories, not any named product.
Honest math.
Free while in beta. If paid tiers arrive, pricing will be published here before anything is charged.
The app
Free while in beta
Dictation, meetings, the library. All of it, no account needed. Updates arrive on their own.
Download for MacModel costs
$0 for raw dictation
On-device transcription costs nothing. Polish and notes can call a model with your own key, billed by your provider, itemized per call. Sign in and an included daily allowance for dictation and a weekly one for meetings work with no key.
What leaves the Mac, exactlyQuestions
Before you download.
What does it ask permission for?
Microphone and Speech Recognition, so it can hear you. Input
Monitoring, so it notices the fn key in any app.
Accessibility, so it can type where your cursor is. Meeting notes
add Screen & System Audio Recording, which is how a Mac hears the
other side of a call; nothing joins the call. Calendar is optional
and read-only. Settings → Permissions shows every one of them,
live.
Does it work offline?
Yes. Apple's recognition runs on this Mac, and on Apple silicon the on-device model produces the final text. Clean-up runs locally, and a meeting still gets a note written on this Mac from what was said. Anything that needs a model there is no route to, such as a question asked of a meeting, waits and says so.
Does it run on an Intel Mac?
Yes, and it is one download for both. The on-device model needs Apple silicon; on Intel, Apple's recognition produces the text, and any provider you add does the polish and the notes.
Which languages?
Whatever your Mac's recognition supports, read from your Mac each time rather than from a list here. The on-device model covers European languages; for the others, Apple's recognition or your own provider produces the text. The languages page has the list a current Mac reports.
What leaves my Mac?
With nothing configured, nothing is sent to us. Add a provider key and the text or audio for that one job goes to that provider, on your account with them. Sign in without a key and the text goes through our relay to Google's Gemini API once, and is not stored. Audio never goes through the relay. The app names the route on every result, and the privacy page lists all three.
Do I need an account?
No. Everything local works without one. Signing in adds hosted model access: an included allowance for polish and writing review by the day, and for meeting notes by the week, with no key of your own. You sign in with Google or with a six-digit code sent to your email. There is no password, and no separate sign-up.
How do updates arrive?
The app checks for a new version when it opens and every hour after, downloads it quietly, and tells you when it is ready. Every release is signed and notarized by Apple, and the changelog says what changed.
Start saying instead of typing
Download. Drag it home. Hold fn. Thirty seconds to your first dictation.
v2.32.9 · macOS 14 Sonoma or newer · Apple silicon and Intel · notarized by Apple
The on-device model needs Apple silicon. Intel uses Apple's recognition and any provider you add.
Release notes by email, if you want them.
Release notes only. Never sold, never shared. Every email says how to leave.