Voice → Text · On-device

Your voice.
Your keyboard.

On-device voice transcription that works as your keyboard. Powered by whisper.cpp. No cloud. No latency. No compromise.

Voxboard listening state
Product VXB-01
Engine whisper.cpp
Processing On-device
Privacy Zero data sent
Platform iOS 17+
01 — How it works

Three steps.
Zero friction.

Voxboard replaces your iOS keyboard with a voice-first interface. Tap, speak, done. Works in every app.

01
Start listening
Open the app and tap Start Listening. The microphone activates in the background, ready when you are.
02
Switch to any app
Open Safari, Notes, Messages — anywhere with a text field. The Voxboard keyboard is always ready.
03
Tap the mic, speak
Hit the microphone button on the keyboard. Speak naturally. Your words appear as text — instantly, privately, on-device.
Voxboard keyboard in Safari
02 — Features

Built for the
real world.

🔒

Fully Private

All processing happens on your device. No audio is uploaded. No transcripts are shared. Your voice stays yours.

⌨️

Custom Keyboard

Works as a system-wide keyboard extension. Use voice-to-text in any app — Messages, Notes, Safari, Slack, anywhere.

🧠

whisper.cpp Engine

Powered by OpenAI's Whisper model, optimized for Apple Silicon with whisper.cpp. State-of-the-art accuracy, locally.

🌍

Auto Language

Automatic language detection across 99+ languages. Or lock to your preferred language for faster results.

📋

Transcript History

Every transcription is saved locally. Swipe up to browse, copy, or reference past recordings anytime.

🎛️

Choose Your Model

Pick from Tiny (75 MB) to Large v3 Turbo (1.6 GB). Balance speed and accuracy for your needs.

03 — Interface

Designed to
disappear.

Minimal. Dark. Focused. The interface stays out of your way so your voice can do the work.

Home — Ready state
01 — Ready to listen
Listening — Active
02 — Listening active
Keyboard in Safari
03 — Keyboard in Safari
Model settings
04 — Model selection
Transcript history
05 — Transcript history
Settings — Privacy info
06 — Privacy & specs
04 — Models

Pick your
engine.

Five Whisper models from ultralight to maximum accuracy. All run entirely on your device.

Tiny
75 MB
Fastest
Small
466 MB
Accurate
Medium
1.5 GB
High accuracy
Large v3 Turbo
1.6 GB
Maximum
05 — Technical specs

Under the hood.

Engine
whisper.cpp
AI Model
OpenAI Whisper (GGML)
Processing
100% On-device
Acceleration
Apple Neural Engine + Metal
Languages
99+ with auto-detect
Model sizes
75 MB — 1.6 GB
Keyboard Extension
System-wide
Always-on mode
Background listening
Transcript storage
Local only
Data sent to cloud
None — zero
Platform
iOS 17+
Framework
SwiftUI + KeyboardKit
🔐

Your voice
never leaves.

No servers. No accounts. No analytics. Every syllable is processed by the Whisper model running directly on your iPhone's Neural Engine. We couldn't access your data even if we wanted to.

Start speaking.

Download Voxboard and turn your voice into text — privately, instantly, everywhere.

Download on the App Store