Tirona turns speech into text anywhere on your phone — in any app, in any text field — without sending your voice to anyone.
Hold the button, speak, let go. The text appears where your cursor already was.
WHY IT IS DIFFERENT
The speech model runs on your phone. Not on a server, not in someone's cloud, not "processed and immediately deleted". By default your voice never leaves the device, so there is nothing to leak, nothing to subpoena, and nothing to trust us with.
If you would rather use a cloud service, you can — Tirona will call OpenAI, Groq, Deepgram or your own server using an API key you supply and pay for. It is off until you switch it on, it is never the default, and the app tells you which engine handled each dictation. We never see the key and we never see the audio.
That also means it works on a plane, on the underground, and on hotel wifi that blocks everything. After the one-time model download, dictation needs no network at all.
WHAT IT DOES
• Dictate into any app — messages, email, notes, search boxes, forms
• Full-accuracy speech recognition, running locally
• Removes "um", "uh" and false starts, so you get the sentence you meant
• Learns how you spell names, from your own notes and your calendar
• Transcribe recordings you already have — voice memos, interviews, calls
• Modes: different settings for different apps, switched automatically
• Optional cleanup pass that tidies rambling into readable prose
FREE, AND NOT THE KIND OF FREE THAT MEANS YOU ARE THE PRODUCT
Everything above is free, with no daily limit, no account and no sign-up. There is no advertising and nothing is sold, because there is nothing to sell — the app never receives your data in the first place.
We can afford this because running the model on your phone costs us nothing.
PRIVACY
• No account, no email, no sign-up
• No advertising and no data sharing
• Usage statistics are off, and off means nothing is counted
• Password fields are detected and refused — Tirona will not type into one
• Calendar and notes access are optional, off until you turn them on, and used only to spell names correctly
Full policy: https://tirona.app/privacy
ACCESSIBILITY
Tirona uses the Android AccessibilityService API to place text into the field you are using. That is what makes dictation work in apps other than its own.
While the dictation bubble is on, Tirona reads which text field has focus, and the contents of that field only to confirm your text arrived. It ignores password fields, and reads nothing while you are not dictating. This information stays on your phone: it is never collected, stored off the device, shared, or transmitted to us or anyone else.
You are asked to agree to this inside the app before the service can be enabled, and you can turn it off at any time under Settings > Accessibility.
REQUIREMENTS
Android 8.0 or later. The speech model is a one-time download of about 630 MB, so use wifi for the first run.
Named for Marcus Tullius Tiro, who invented shorthand in 63 BC to write down Cicero's speeches at the speed they were spoken.
Hold, speak, release. Speech becomes text on your phone, not on a server.