Dictate anywhere.
Nothing leaves your device.
opennib turns your voice into text in any app — on a model that runs entirely on your phone or computer. No accounts, no cloud, no tracking. Hold a key, speak, release.
Voice to text, the way it should work.
Built for people who type all day and care where their words go.
Hold to talk
Press your hotkey (or the keyboard mic), speak, release. The transcript drops in right where your cursor is.
Runs on your device
The speech model lives on your machine. After a one-time download, opennib does its work without the network.
Completely free
No subscriptions, no in-app purchases, no account — and it always will be. Free and open source, forever.
Everywhere you type
A system-wide keyboard on iPhone and Android, a menu-bar app with a push-to-talk key on macOS, Windows and Linux.
Private by design
No telemetry, no analytics, no servers. We genuinely can't see what you say — and the code proves it.
Open source
MIT licensed. Inspect every line, build it yourself, or contribute on GitHub.
Everything happens on this device.
Audio, transcripts, and your dictionary never leave your phone or computer. No telemetry, no cloud, no servers — opennib is offline the moment the model finishes downloading.
Get opennib, free.
Completely free — no subscriptions, no in-app purchases, no account. One-time model download (about 470 MB on desktop, 75 MB on mobile), then it’s yours offline.
Private dictation, answered.
The things people ask before switching from cloud dictation tools.
Does opennib send my voice to the cloud?
No. Speech recognition runs on your own device with the Whisper model. Audio is transcribed locally and discarded the moment the text exists. There are no opennib servers and no account, so there is nothing to send and nowhere to send it.
Is opennib really free? What's the catch?
There isn't one. opennib is free and open source under the MIT license: no subscription, no in-app purchases, no "pro" tier. It exists because private, offline dictation shouldn't cost a monthly fee.
Does it work offline?
Yes. The first run downloads a speech model (about 470 MB on desktop, 75 MB on mobile). After that, dictation works on a plane, in a tunnel, or with Wi-Fi off.
How is this different from Wispr Flow, Apple Dictation or Google voice typing?
Those services send your audio to a server. opennib is the offline, open-source alternative: the same hold-to-talk convenience in every app, with the speech model running on your hardware and the code public for anyone to inspect.
Which languages can I dictate in?
opennib uses multilingual Whisper models, so it handles dozens of languages including English, Vietnamese, Spanish, French, German, Japanese and Chinese. Choose one in settings, or let it detect the language automatically.
What does it run on?
macOS on Apple Silicon today, with Windows and Linux builds in preview. On iPhone, a system-wide keyboard puts dictation in any text field; on Android, an input method does the same. Everything is built on the same open-source core.