Say Kubernetes. Get Kubernetes.
A general speech model hears “Kubernetes” and writes “hubernitz”, and turns your camelCase identifiers into word soup. Phrasora leans on a dictionary that learns your stack — so the jargon you actually say comes back spelled the way you mean it, fast, in any app, including your terminal.
Free for 2,000 words a week, no card. Mac, Windows, iPhone, Android, and the web.
It learns the words your model never met.
Phrasora gives the cloud transcription a short, focused hint drawn from your own dictionary — service names, library names, the identifiers you say a hundred times a day. The hint is evidence-gated, so a term only rides along when there’s a real sign you spoke it. It nudges toward your words without forcing one you didn’t say.
It also reaches inside fused tokens, so a punctuated or camelCase term can match a name you’ve taught it. Correct a term once and that fix follows you across every device, with time to undo every learned moment.
A term you spelled right stays exactly as you said it — it won’t “correct” an identifier you already got right.
let’s deploy this to hubernitz before the postgress migration
PhrasoraLet’s deploy this to Kubernetes before the Postgres migration.
Spell out anything it hasn’t seen.
For a brand-new repo name or a private acronym, say it and then spell it — “the service is named, A-T-L-A-S” — and the text reads “Atlas”, spelled your way. The letters always win over what was heard, even inside a longer phrase. Acronyms and flags you genuinely dictate stay exactly as you said them.
It only steps in for a real spell-out — when you’re not spelling something out, it leaves your words alone.
run the migration on the, A-T-L-A-S, service tonight
PhrasoraRun the migration on the Atlas service tonight.
One hotkey, every app, your terminal too.
Phrasora is a system-wide dictation layer, not a box you type into. On Mac and Windows, hold your hotkey and clean text pastes at your cursor in your editor, a terminal, a pull-request description, a chat — anywhere. The same dictionary and history follow you to iPhone and Android.
And it’s quick: a median of 871 milliseconds from releasing the key to text on screen, with the 90th percentile still under a second and a half. The same full-accuracy cloud model runs on every platform.
Three steps to dictating anywhere.
Download for your platform
Grab the Mac or Windows app from the downloads page (signed, with built-in updates), or install the iPhone or Android app once it's live in the store.
Grant the permissions it needs
On Mac, allow what it asks for — microphone, speech recognition, and Accessibility so it can insert text in any app. On Android, grant Accessibility — that single permission powers both the floating button and text insertion, and Android 13 and later may also ask to send notifications. On Windows it just runs, though it may prompt for microphone access on first use.
Set your hotkey and dictate
On Mac and Windows, pick a global hotkey, hold it, and speak — clean text lands at your cursor anywhere. On iPhone, enable the Phrasora keyboard with Full Access and switch to it, then dictate from any field.
On Android, granting Accessibility is the one permission that powers both the floating button and text insertion.
The questions developers ask first.
Will it actually get technical terms right?
A general speech model has never heard "Kubernetes", your repo's name, or a camelCase identifier. Phrasora gives the cloud transcription a focused hint from your own dictionary, so terms you use come back the way you spell them — but only when there's evidence you actually said them, so it never forces a word in. Add a term once and the correction follows you across every device.
How do I dictate something it has never seen?
Say it, then spell it. "camelCase, C-A-M-E-L, case" or "the repo is named, P-H-R-A-S-O-R-A" — the letters always win over what was heard, even mid-sentence. Acronyms and codes you genuinely dictate are left exactly as you said them.
Does it work in a terminal?
Yes. On Mac and Windows, Phrasora pastes clean text at your cursor in whatever app is in front of you — your editor, a terminal, a browser, a chat client. It's a system-wide dictation layer, not an in-app box, so the same hotkey works everywhere you write.
How fast is it, really?
Measured end to end, from key release to text at your cursor, the median is 871 milliseconds and the 90th percentile is 1,149 milliseconds — wall-clock on the production pipeline. Speech goes to cloud transcription, the AI cleanup runs, and the finished text is pasted in.
Dictate your next commit message.
Free for 2,000 words a week, no card required. Teach it your stack, hold the hotkey, and watch the jargon land right the first time — in your editor, your terminal, anywhere.