Everything you want to know before you download.

airtxt installs an iOS keyboard with a mic button. Tap it, speak, and the airtxt app records and transcribes your speech, cleans it up with AI, and the keyboard inserts the finished text wherever your cursor is — Mail, Slack, Notes, Messages, any app with a text field. The first tap opens the airtxt app for a second to switch the microphone on (Apple requires this); after that, dictating is one tap.

No. You tap the globe icon on your current keyboard to switch to airtxt when you want to dictate, then switch back when you're done. Or set airtxt as your default keyboard — it has a full typing layout, so the mic stays one tap away without any switching.

It's the best way to use it. The airtxt keyboard has a full typing layout, so you can type normally and dictate from the same keyboard. With airtxt as your default and a dictation session running, speaking into any app is a single tap.

It removes filler words (um, uh, like), false starts, and repetitions. It adds punctuation, fixes capitalization, and formats lists or paragraphs when the content calls for it. You speak naturally; the output reads as if you typed it carefully.

No. In a quiet setting, airtxt picks up soft speech — even a whisper. Useful in an open office, a library, a shared apartment, or anywhere you'd rather not narrate your email out loud.

Any iOS app that accepts keyboard input — Mail, Messages, Slack, WhatsApp, Notes, Notion, Safari, Chrome, Google Docs, Word, LinkedIn, Twitter/X, and so on. The one exception is password fields: iOS blocks every third-party keyboard there and switches you to Apple's keyboard automatically.

Yes. Pin any of 13 languages in Settings — Spanish, French, German, Portuguese, Italian, Japanese, Korean, Mandarin, Hindi, Arabic, Russian, and English (US or UK) — or leave it on Automatic to use your phone's language. The AI cleanup works in the language you spoke; it never translates.

Turn on Multilingual in Settings and airtxt detects what you're speaking as you speak it — including switching languages mid-sentence, with no button press. It covers English, Spanish, French, German, Hindi, Russian, Portuguese, Japanese, Italian, and Dutch. Auto-detection is a cloud-mode feature; in on-device mode, dictation uses your phone's language.

In on-device mode, transcription runs entirely on your iPhone using Apple's Speech framework. Your audio is never sent to any server — not to airtxt, not to Deepgram, not anywhere. The transcribed text is then sent to airtxt's servers over an encrypted connection for AI cleanup. It is the most private way to dictate: the audio itself never leaves your hands.

Cloud mode uses Deepgram Nova-3 and gives you higher accuracy, especially for technical terms, proper nouns, and accents. It is also where Multilingual auto-detection runs. Use it when accuracy matters more than strict data locality — for most users, most of the time.

No. airtxt never stores audio anywhere. In on-device mode, audio never leaves your phone at all. In cloud mode, audio is streamed to Deepgram for transcription only. Your dictation history — the resulting text — lives on your phone, capped at your most recent 200 dictations.

On-device mode keeps audio on your iPhone — nothing is ever recorded to a server. The transcribed text is sent to airtxt's servers, encrypted, for AI cleanup, and airtxt does not store it. Whether that fits your compliance requirements (HIPAA, privilege) is your call to make — we publish exactly what leaves the device so you can make it.

Typing works without Full Access. Dictation needs it — Full Access is how the keyboard receives your finished transcript from the airtxt app. The keyboard itself never records audio and never sends your keystrokes anywhere. Setup walks you through the one switch to flip.

Paste the meeting link into the airtxt app and the bot joins your Zoom, Google Meet, or Microsoft Teams call as a participant. It records, labels each speaker, and produces a clean transcript.

A full speaker-labeled transcript, plus a follow-up email drafted in your voice — it leads with the outcomes, then next steps with owners, ready to review and send.

The bot joins as a named participant, so it appears in the participant list and every attendee can see it. Whether it gets in is the meeting organizer's call: depending on the platform and meeting settings, the host may need to admit it from the waiting room, and the host can remove it at any point. airtxt works whenever the bot is allowed in the meeting.

You get a transcript for the time the bot was actually in the call. If it was never admitted, there is no recording and no transcript — airtxt cannot capture a meeting it wasn't allowed into. Only the minutes the bot actually spent recording count against your meeting-hour allowance.

Yes. You can use them independently. The Pro plan covers both.

A list of words the transcription engine should always spell a specific way — product names, technical acronyms, people's names, medical terms, legal citations, anything speech engines typically mangle. You add them once and they apply in both on-device and cloud modes.

Yes. Custom vocabulary is included on the Free plan. Most voice apps charge extra for this.

There is no enforced limit during beta. We will set a reasonable ceiling before general release and it will remain generous.

Unlimited on-device dictation, with AI cleanup capped at 2,000 words a week. Custom vocabulary is included. There is no meeting bot on Free and no cloud transcription — enough to evaluate airtxt's core dictation quality before committing.

Unlimited dictation with AI cleanup on a generous fair-use allowance, cloud transcription (about 300 minutes a month), 5 hours of meeting bot time a month, follow-up email drafts, and meeting history with search — all for $9.99/mo or $99.99/yr. Need more meeting time? Top up in five-hour packs, no surprise overage.

When paid plans launch with the App Store release, iOS subscriptions will go through Apple In-App Purchase (StoreKit). You'll be billed by Apple, manage payment in your Apple ID settings, and can cancel there at any time. During the beta, everything is free — there is no billing yet.

Beta users get full Pro access free during the TestFlight period. Pricing activates when the app ships on the App Store.

Pro+ is $19.99/mo and is built for meeting-heavy weeks: 15 hours of meeting bot time a month with priority cloud transcription, plus everything else Pro includes. You can top up more meeting hours anytime.

airtxt requires iOS 17 or later. On-device mode with the latest Apple Speech engine (SpeechTranscriber) requires iOS 26. On iOS 17–25, on-device mode uses the SFSpeechRecognizer framework — slightly lower accuracy but fully private.

Not yet. A desktop app is on the roadmap after the iPhone app ships. For now, your dictation history, meeting transcripts, and settings all live in the iOS app.

Yes. airtxt uses whichever audio input the system routes to. AirPods, EarPods, Bluetooth headsets, and the built-in microphone all work.

The dictation fails with an error and nothing is inserted — you can speak it again. Switching to on-device mode, which works with no connection at all for transcription, is one tap in Settings.