Push to talk
Hold one key, speak, release. Text lands where your cursor is — Slack, Gmail, your editor, anywhere you can type.
Hold ⌥. Say it. It's typed.
Semoru turns your voice into clean text in any app on your Mac — in any language, even Bangla and English mixed in one sentence. Fillers removed, punctuation placed, your words kept.
free · 3,000 words a week · macOS on Apple silicon · beta
the whole interface — a 22-pixel bar at the bottom of your screen
bangla + english · একসাথে
The first dictation that speaks your mix
Code-switching is how millions actually talk. Say the sentence the way you'd say it — Semoru writes each word in its own script, or romanises everything to Banglish if you prefer. Set the output to English and it translates as you speak.
you say
"amar meeting ta cancel hoye geche, kal call korbo"
semoru types
আমার meeting টা cancel হয়ে গেছে, কাল call করব।
features
Everything between your voice and the page
On this Mac
Transcription runs on your computer. Works offline. Your voice never leaves the laptop.
Cleanup you can see
Fillers, repeats and self-corrections removed. History keeps raw and clean side by side — one key brings the raw back.
Any language in, English out
Set the output to English and speak whatever you like — Semoru translates as it types.
Per-app style
Casual in Slack, formal in Gmail, untouched in your code editor. It also learns how you write — subtly.
Spoken commands
"new line", "new paragraph", "scratch that" — structure and undo, by voice, in English or Bangla.
Snippets
Say "my address" and the full address appears. Save anything you repeat but never want to spell out loud.
Dictionary that learns
Fix a word once and Semoru remembers the spelling — names, brands, jargon come out right next time.
History and stats
Searchable, local, yours. Words per day, speed, streak — computed on your machine only.
privacy
Built like it's listening to everything — because it is
A dictation tool hears your work, your names, your numbers. That demands more than a policy page. It demands architecture.
01
On-device by default
On Apple silicon, speech becomes text on your Mac. Airplane mode? Still works.
02
Zero retention, in code
When the server is used, audio streams into memory, is transcribed, and the buffer is freed. No disk, no logs. Transcripts live only on your machine.
03
No third-party AI vendors
Open models on our own servers. Your words are never routed to an AI company's API.
04
You can watch it
A live log in the app shows every request: what was sent, how large, which region. It's usually empty.
05
Screenshots — never
Semoru cannot see your screen. Not a setting. Not a toggle. Absent.
pricing
Free to start. Fair when you're hooked.
Free
$0
- 3,000 words a week
- Every language, mixed speech
- On-device transcription
- History, dictionary, snippets
no card, no trial clock
Pro
$8 /month
- Unlimited words
- Translate as you speak
- Priority cleanup
- Priced fairly by country
billing opens after beta — beta testers dictate free
Lifetime
$149 once
- Everything in Pro
- Forever, one payment
- First in line for new platforms
limited at launch
download · beta
Two minutes to your first dictation
Download Semoru for Mac · 9 MB
Requires macOS 12+ on Apple silicon (M1 or newer).
Open it the beta way
Drag Semoru to Applications. First launch only: right-click → Open → Open. If macOS still objects: System Settings → Privacy & Security → Open Anyway. We're a beta without Apple's $99 stamp yet — the app is signed and updates are verified cryptographically.
Let it set up
Allow the microphone and Accessibility when asked, sign in with your email, and download the on-device model (570 MB, once).
Hold ⌥ and talk
That's the product. Updates arrive by themselves from here on.
Windows or Linux?
Same app, same privacy. Leave an email and you'll get the first build.
You're on the list. One email when it ships — nothing else.