Sayboard Offline AI Keyboard: iPhone Setup & Cost

Which Sayboard are you looking for?

This page covers Sayboard: Offline AI Keyboard by Stanislav Cherniavskyi, App Store ID 6760622185, linked from sayboard.app. Check the developer and app ID before following setup or pricing advice.

There is also an iPhone app called Sayboard: Voice Keyboard by Joy Madhu, ID 6795650236, with a different subscription model. Separately, the Android Sayboard project by ElishaAz uses Vosk. Neither listing should be used to infer this offline iPhone app’s price or privacy behavior.

Price: paid download or free source code?

The U.S. App Store price is $7.99, checked September 6, 2026. The listing describes a one-time purchase without subscriptions or in-app purchases. Local prices can differ; verify the purchase sheet before buying.

The developer also publishes the source under GPL-3.0. Building it yourself is a separate route that requires a Mac, Xcode, signing configuration, and developer experience. Open source does not mean the ready-to-install App Store version is free.

What to set up before going offline

Install the correct app and complete its model download while connected. Enable Sayboard under iPhone Settings → General → Keyboard → Keyboards → Add New Keyboard, then switch to it in an ordinary text field. Follow its onboarding for microphone and keyboard permissions.

The site lists separate speech and text models. Its published examples range from roughly 29–466 MB for speech and 551 MB–2.1 GB for text cleanup, with larger text models subject to device memory limits. Treat those as examples, not a complete current catalog or the total install size.

Our suggested check: download one supported speech model, dictate a disposable sentence in Notes, then repeat in airplane mode. If you also need rewriting, test that separately after downloading its text model. These are verification steps for your device, not a hands-on result from us.

Why does it switch to the app?

The developer’s FAQ says Sayboard briefly opens its containing app to activate the microphone before returning to the keyboard. It describes this as a session-start step and offers a session-duration setting. Expect that handoff rather than assuming dictation always starts entirely inside the keyboard.

If you cannot dictate, first check that a model finished downloading and that microphone permission is enabled. If a larger text model is unavailable, consult the device-memory requirements. Speech recognition and rewriting are separate stages, so a working microphone alone does not confirm that both are ready.

What “offline” means for privacy

The privacy policy says speech, recordings, and transcription history stay on the device, with no analytics or accounts. It makes an explicit network exception for model downloads. Apple handles the purchase separately. These are developer disclosures; we have not audited the released app.

Local processing does not mean there is no saved history. The policy describes clearing recordings, transcripts, and downloaded models in settings. Review those controls if you share the device.

The source repository also describes a voluntary problem-report feature that prepares an email or GitHub issue with app version, iOS version, and a device identifier. Inspect a report before sending it; choosing to share diagnostics is different from automatic telemetry.

Who should consider it?

Our recommendation: consider it when local processing and avoiding a recurring bill matter enough to accept model downloads and device-dependent performance. A smaller phone or a need for effortless setup may change that tradeoff.

Before relying on it, compare a short message, an unfamiliar name, and a longer paragraph against your current keyboard. Check meaning as well as spelling after cleanup. We have not measured accuracy, latency, heat, or battery use, and do not rank it above cloud keyboards on those measures.