Your voice, your text,
your privacy.
TypeSay converts your speech to text right on your machine. No cloud, no accounts, no data collection. Just hold a hotkey, speak, and release. Your words appear wherever your cursor is.
Three steps. Zero complexity.
No setup wizards, no account creation, no configuration menus. TypeSay works the moment you install it.
Hold your hotkey
Press and hold the keyboard shortcut. The default is Ctrl + Shift + Space
Speak naturally
Talk at your normal pace. The floating visualizer orb animates in real time so you know TypeSay is listening.
Release to type
Let go of the hotkey. Your words are transcribed locally and inserted at your cursor position in under 3 seconds.
Built different. On purpose.
Every design decision serves one goal: give you the best speech-to-text experience without compromising your privacy.
100% Private
Your audio never leaves your machine. No cloud processing, no telemetry, no analytics. Everything runs locally through Whisper AI.
One-Time Purchase
Pay once, own it forever. No monthly fees, no annual renewals, no surprise charges. $89.99 and it's yours.
Works Everywhere
System-wide hotkey works in any application. Email, Slack, VS Code, Word, your browser, terminal. If it has a text field, TypeSay fills it.
Lightning Fast
The Base model transcribes 10 seconds of speech in under 3 seconds. GPU acceleration makes it even faster with CUDA, Metal, or Vulkan.
Everything you need. Nothing you don't.
Your data never leaves your machine
TypeSay runs OpenAI's Whisper model directly on your hardware using whisper.cpp. There are no API calls, no cloud uploads, and no server-side processing. Your audio is captured, transcribed, and discarded in memory. Nothing is saved. Nothing is sent.
In a world where AI tools demand your conversations, your files, and your data as the price of admission, TypeSay takes the opposite approach. Your voice is yours alone.
Four models. Your choice of speed vs. accuracy.
Choose from Tiny (75 MB) for instant transcription, Base (142 MB) for everyday use, Small (466 MB) for high accuracy, or Medium (1.5 GB) for the best results. Models download once from HuggingFace and run entirely offline after that.
Switch models anytime from the settings panel. No restart required.
Native performance. Tiny footprint.
TypeSay is built on Tauri 2.0 with a Rust backend and SolidJS frontend. The installer is approximately 5 MB. Memory usage when idle is minimal. When recording, the Whisper engine activates only for the duration of your speech.
GPU acceleration is available through CUDA (NVIDIA), Metal (Apple Silicon), and Vulkan (cross-platform).
How TypeSay stacks up
See how TypeSay compares to built-in OS dictation and cloud-based transcription services.
| Feature | TypeSay | Windows Dictation | macOS Dictation | Cloud Services |
|---|---|---|---|---|
| Fully Offline | ✓ | ✗ Cloud | ✓ On-device | ✗ Cloud |
| No Account Required | ✓ | ✓ | ✓ | ✗ |
| No Subscription | ✓ One-time | ✓ | ✓ | ✗ |
| Cross-Platform | ✓ Win/Mac/Linux | ✗ | ✗ | ✓ Web only |
| Open Source | ✓ MIT | ✗ | ✗ | ✗ |
| GPU Acceleration | ✓ CUDA/Metal/Vulkan | ✗ | ✗ | ✗ |
| System-Wide Paste | ✓ | Limited | Limited | ✗ |
| Multiple AI Models | ✓ 4 tiers | ✗ | ✗ | ✗ |
| Visual Feedback | ✓ Animated orb | ✗ | ✗ | ✗ |
Type at the speed of thought.
One-time purchase. No subscriptions. No cloud. Just you, your voice, and your machine.