Processed, never stored
Speak instead
of type.
TypeMic turns speech into text in about a second, and keeps none of it. No voice logs, no training on what you say.
Windows, iPhone and Android are on the way, get notified.
How privacy works ↗TypeMic in motion, a sixty-second film
Sixty seconds, start to finish, hotkey, speech, and the words landing at the cursor.
The one-second proof
About a second, every time.
About a second from voice to text, at any dictation length.
Built on CloudflareISO 27001 · SOC 2 Type II networkWhere it types
Dictate where you already work.
Terminal, browser, IDE, chat, the words land at the cursor.
- hold
Press your shortcut
One global hotkey. Nothing to open, no window to switch to.
- speak
Say the words
A minimal on-screen waveform confirms it’s listening.
- release
Text appears
About a second later the words are at your cursor, terminal, browser, IDE, or chat.
Privacy, stated plainly
Spoken, typed, forgotten.
Two modes, side by side, no hedging, no burying.
On by default
Cloud
Transcription runs through our own endpoint on Cloudflare's global infrastructure. Audio is processed, never stored.
The engine
Predator
Our own transcription engine behind that endpoint, on Cloudflare's global infrastructure, load-balanced across numerous nodes worldwide, so every user is served close to where they are. Engineered for speed, security, and privacy; benchmarked at 0.9 seconds from release to text.
Cookieless analytics. No ad trackers or analytics cookies.
Read the privacy page →Compare TypeMic with eight dictation apps →
Questions
Asked before every download.
What does TypeMic cost?
Sign in once with Google and your first 7 days are unlimited, free, no card. After that you keep 2,000 dictated words every week, free forever. TypeMic Pro unlocks unlimited dictation on every device you link, $12/month or $114/year. The iPhone and Android keyboards are in development.
Does dictation work in every application?
TypeMic types at the cursor in whatever app has focus, terminals, browsers, IDEs, chat clients. A global hotkey starts and stops recording, and an on-screen waveform shows it is listening. Typing goes to the focused window, and behaviour can vary between desktop environments.
Does my voice ever leave my computer?
It is never stored. Each utterance goes to typemic.com over an encrypted connection, is transcribed by TypeMic's own Predator engine on Cloudflare's global network, certified to ISO 27001, ISO 27701, SOC 2 Type II and PCI DSS, typed at your cursor, and discarded immediately: no voice logs, no training on your words, and the app shows a notice at every start so nothing about the mode is silent.
How is TypeMic different from built-in dictation?
Built-in dictation belongs to the app or the operating system that offers it. TypeMic is one global hotkey for whatever has focus, terminals, browsers, IDEs, chat, with a cleanup pass that punctuates and capitalizes, and nothing is stored: audio and transcripts are processed and discarded. On Linux there is no built-in dictation to compare with; on a Mac, Apple's dictation stays available beside TypeMic.
Engineering
Privacy by design, from the first line of code.
Written in Rust, for performance, security, and quality. Engineered by cybersecurity leaders with 20+ years securing enterprise and mid-market organizations, the same engine and the same privacy engineering on every platform we ship.
engineered, not assembled
▸ written in Rust
▸ S.O.L.I.D. principles
▸ ISO 27001 compliance
▸ 20+ yrs enterprise securityArchitected, secured, and maintained by the founders of CyberLink Security. As a matter of fact, we used TypeMic to build TypeMic.
Infrastructure
Built on Cloudflare, stated with their numbers.
The Predator engine and every TypeMic service run on Cloudflare’s global network. The figures below are Cloudflare’s own, each linked to its source. The 0.9 seconds is ours, measured.
Speed
Within 50 ms of 95% of the world’s internet users
Most within 20 ms, across 330+ cities. On that network the Predator engine is benchmarked at 0.9 seconds from release to text, at any dictation length.
AI inference
Inference in 200+ cities worldwide
Workers AI runs the Predator engine close to the user, load-balanced across numerous nodes, so every dictation is served near where it was spoken, with the same engineering everywhere.
Security
ISO 27001, ISO 27701, SOC 2 Type II, PCI DSS
Cloudflare’s network holds those certifications. TypeMic’s administrative surfaces sit behind Cloudflare Zero Trust at the edge, so unauthenticated requests never reach the application, and TypeMic itself is engineered to ISO 27001 compliance by the founders of CyberLink Security.
Protected and accelerated by Cloudflare, with security engineering by CyberLink Security and RaptorLabs.
Platforms
Linux and Mac today. Three planned.
TypeMic ships on Linux and macOS now, one codebase, the same engine, signed on both. Windows, iPhone, and Android are planned; pick yours and we’ll email you the day it opens. One email, no spam.
Pricing
Free to start. Unlimited when you need it.
Free
$0
Free Google sign-in, no card, first 7 days unlimited.
Then 2,000 dictated words every week, free forever.
Get TypeMic
Speak instead of type.
TypeMic turns speech into text in about a second, and keeps none of it.