Say it.
It’s written.
Press a hotkey and talk the way you think. Ormia turns it into clean, punctuated text at your cursor in Slack, Mail, Docs, your code editor or any other app.
Private beta for Mac, Windows and Linux. Free on your own machine, on open-weight Qwen3 models.
Fillers gone, the correction applied, the list formatted. Nothing you didn’t say was added.
Talk like a person. Get text like an editor wrote it.
Dictate anywhere
One hotkey starts and stops. Text is pasted where your cursor is, and your clipboard is put back afterwards.
Cleans up after you
Drops the ums and false starts. “Tuesday, no wait, Wednesday” becomes Wednesday. Lists come out as lists.
Edit by voice
Select text and say “make this shorter” or “translate to Spanish”. With nothing selected, say what to write.
The right tone per app
Relaxed in Slack and WhatsApp, polished in Mail and Outlook, faithful everywhere else. You set the rules.
Knows your words
Add names, products and jargon once. Ormia spells them your way every time.
Never lose a word
Every dictation is kept in a local history. If a paste fails, it’s one click away.
Three keystrokes’ worth of effort.
-
01
Press the hotkey
⌘ ⇧ Space from any app. A small pill shows Ormia is listening, and focus never leaves what you’re doing.
-
02
Talk naturally
Ramble, correct yourself, say “new paragraph”. Qwen3-ASR transcribes, then Qwen3 tidies it in the tone that suits the app.
-
03
Press again. Done.
Finished text appears at your cursor. Press Esc instead to throw the take away.
Built for how the world actually talks.
Thirty languages and twenty-two Chinese dialects. Pick one, or let Ormia detect it.
Your voice, your machine. You choose where it goes.
- On your computer · default
- Both models run locally. Audio never leaves the machine, and Ormia works offline.
- On your own server
- Point Ormia at a GPU box in your own cloud or data centre. Data stays inside your walls.
- Hosted
- No setup on older laptops. Audio goes through OpenRouter to the provider running the same open-weight models, and Ormia doesn’t keep it.
The microphone is open only while you’re dictating. API keys are encrypted with your system keychain.
No app? Send audio, get clean text.
The same transcription and cleanup as an API, for your own product, scripts or phone shortcuts. It speaks OpenAI’s transcription format, so most tools switch by changing one URL.
Free accounts get 30 minutes of audio a month, Pro 300, Teams 600 per seat.
Request an API keycurl https://ormia.ai/v1/audio/transcriptions \
-H "Authorization: Bearer $ORMIA_API_KEY" \
-F file=@note.m4a \
-F style=formal
{ "text": "Let’s move the launch to Friday." }Free where it runs free.
Pro
$12 / month, or $108 a year
- Hosted models, nothing to install
- Faster on any laptop
- Sync dictionary and styles across devices
- 300 minutes a month on the API
Teams
$16 / seat / month, billed yearly
- Run on your own servers
- Shared team dictionary
- Central settings and rollout
Questions
When can I download it?
Ormia is in private beta. Leave your email below and we’ll send the download link, or an API key, as each group opens.
Which computers does it run on?
Mac, Windows and Linux. For fully local use, an Apple Silicon Mac or a PC with a recent GPU gives the best speed. Older machines can use the smaller speech model or hosted mode.
Will it answer my question instead of typing it?
No. Ormia only rewrites what you said. If the cleanup model ever drifts from your words, Ormia falls back to the plain transcript.
Which models does it use?
Qwen3-ASR for speech and Qwen3 for cleanup. Both are open-weight models released under Apache-2.0, so they can run anywhere you choose.
Does it work offline?
Yes, in local mode. Once the models are downloaded, no internet connection is needed.
Stop typing everything.
Ormia is in private beta. Leave your email and we’ll send your download link or API key as soon as your spot opens.