Free & Open Source

You speak faster than you type.
But you write better than you speak.

Mesmer learns the gap between the two — and closes it.
Speak a message. Get back something that sounds like you wrote it.
Not a transcript. Not AI polish. You.

Download free — macOS See how it works

On-device. Private. Open source.

Every voice tool today gives you two choices. Neither is right.

What you get from dictation tools
"yeah so i was thinking we should probably push the meeting to like thursday or friday because i have that call with the client on wednesday and it runs long usually so yeah"
Unprofessional. You'd never send this.
What you get when you "clean it up with AI"
"I wanted to reach out regarding the possibility of rescheduling our upcoming meeting to either Thursday or Friday, as I have a prior commitment on Wednesday that may run over."
Synthetic. Nobody talks like this. They know you didn't write it.
What Mesmer gives you
"Can we shift our meeting to Thursday or Friday? I've got a client call Wednesday that tends to run long."
You. At your best.

It knows your writing before you send the first message.

Mesmer doesn't start cold. Before you speak a word, connect your Gmail, paste a few Slack messages, drop in an email thread. Mesmer reads how you already write — your rhythm, your register, your shortcuts — and builds your voice profile from that. The first message you dictate already sounds like you typed it.

"Won't it take weeks to learn my style?"

It knows who you're talking to. Not just how you talk.

The message you send a client is not the message you send your co-founder. Mesmer reads the context — the app you're in, the thread you're replying to — and adjusts automatically. Same voice. Different register. You don't configure anything. It figures it out.

"Won't it just use one generic tone for everything?"

Every message makes it sharper. You'll notice by day three.

Every time you edit a Mesmer output — keep something, change something, delete something — it learns. It notices your patterns without asking. After a week, it rarely needs correction. After a month, you stop thinking about it. You just speak and send. That is what a tool that actually knows you feels like.

"AI tools always stay generic no matter how long I use them."

Works in every app. Stays out of your way.

A floating toolbar that follows you — Slack, Gmail, WhatsApp Web, Notion, Xcode, anywhere you type. Place your cursor. Speak. Done. No copy-paste. No switching windows. No workflow broken.

"These tools only work in some apps."

Everything stays on your Mac. No accounts. No servers. No exceptions.

Dictation runs on Apple's Speech framework. Rewriting runs on Apple's FoundationModels. Nothing touches a cloud. No API key. No subscription that needs your email address. Your words stay yours — completely, technically, permanently.

"If it's AI-powered, my messages must be going somewhere."

Stop editing your own voice
into messages.

Download Mesmer free. Open source. Runs entirely on your Mac.

macOS 26+ required Apple Silicon & Intel MIT licensed