say it once
text lands where you’re already typing.
Dictate from the VocaPhone keyboard and keep going in Messages, Notes, email, or any other text field. No copy-and-paste detour.
private speech-to-text, running on your phone
voice typing that never has to leave your phone.
27 language choices · support depends on model · gateway optional
Run speech-to-text entirely on your phone, or choose VocaGateway when you want more models and shared compute on hardware you control.
say it once
Dictate from the VocaPhone keyboard and keep going in Messages, Notes, email, or any other text field. No copy-and-paste detour.
keep it on the phone
Turn on on-device transcription and VocaPhone runs the model right on your iPhone or Android phone. After the model is downloaded, dictation works offline. A gateway is available, but never required.
sound like yourself
Choose from 27 languages plus automatic detection, pick a writing style, and teach VocaPhone the names and phrases you use most. Available languages are filtered to what your selected model can actually transcribe.
Download a model that fits your phone. Every file is integrity-checked before it runs.
Use the voice keyboard and speak naturally in your own rhythm.
Audio becomes text locally, then the successful recording is removed.
tiny honest detail: iOS keyboards cannot use the microphone, so VocaPhone records in its companion app and returns the finished text to the keyboard. The model still runs on the iPhone.
These are unedited VocaPhone screens from Android. On-device mode is visible, selectable, and usable without configuring a gateway.
Keep everything on your phone for the simplest offline setup. Or run VocaGateway on your Mac, Linux machine, or home server when you want broader model and engine choices, more processing power, or one shared transcription setup for your devices.
The model and audio stay on your phone. After the model download, transcription works offline.
Your phone sends the recording to the gateway you configured. It runs a local speech engine and returns the transcript.
One privacy stance, built around the way each phone actually works.
WhisperKit/Core ML models run on the iPhone with no network or gateway needed for transcription.
iPhone setup ↗Native whisper.cpp and sherpa-onnx models run inside the Android phone and insert text directly.
get the beta ↗VocaPhone is free, open source, and built to run speech-to-text directly on your phone. No gateway is required. No account. No subscription. No analytics SDK quietly keeping score. VocaGateway is the optional, self-hosted path for more models or shared compute.
read the code on GitHub ↗No. In on-device mode, recording and speech-to-text both happen on the phone. After a transcript is safely stored, the successful audio file is deleted. If you deliberately choose gateway mode, audio goes to the gateway you configured. Use a trusted LAN, encrypted private network, or HTTPS to protect it in transit.
No. Download an on-device model once and VocaPhone can transcribe without a gateway, account, or network connection. If you want larger models, more engines, or shared hardware, you can self-host VocaGateway on a Mac, Linux machine, or home server and pair your phone by QR.
VocaPhone offers 27 language choices plus Automatic. Actual support depends on the speech-to-text model you select, and the app hides choices that the active model cannot handle.
It works wherever the operating system allows a custom keyboard and editable text field. Password fields and some security-sensitive screens may force the system keyboard instead.
Apple does not let keyboard extensions access the microphone. The companion app records, shares the finished session with the keyboard, and lets the keyboard insert the transcript at your cursor.
After setup, private on-device dictation needs only your phone and a downloaded model it supports. Android has a public beta. iPhone currently requires a Mac, Xcode, Apple signing, and a physical iPhone for the initial source build. Gateway hardware remains optional.
Yes. VocaPhone is free and AGPL-3.0 licensed. There is no VocaPhone account, subscription, or paid cloud speech service.
VocaPhone is in beta. Here is exactly what each platform needs today.
public beta
Download vocaphone.apk and verify it with the
SHA256SUMS.txt included in the same release.
build from source
There is no App Store or TestFlight build today.
ready when you are
Download a model to your phone and turn speech into text without sending your voice anywhere.
no gateway required · free & open source · AGPL-3.0