Version 2.0.0 Now Available

Speak your mind.
AI does the rest.

Free, local AI-powered dictation that turns your voice into text instantly. No cloud, no subscription, no account required. Just download and speak.

Download for Windows View on GitHub
100%Offline
20+Languages
5Model Sizes
$0Monthly Fee
MITLicense
Changelog

What's New in v2.0.0

A complete overhaul that makes dictating faster, smarter, and more powerful than ever before.

Local AI Cleanup
Automatically fix grammar, remove filler words, and rewrite in chosen tones. Powered by Gemma 4 E2B running fully on-device with zero cloud calls.
20 Languages
Dictate natively in 20 supported languages including Spanish, Italian, Japanese, Arabic, Chinese, Russian, and more.
GPU Acceleration
CUDA-capable GPUs now run Whisper transcription for much faster results. Automatically falls back to CPU if no compatible GPU is detected.
Select Your Model
Choose from 5 Whisper sizes Tiny through Large-v3-turbo directly from Settings to balance speed against accuracy for your hardware.
Session History
Every session saves to a searchable local database. View, copy, paste, or delete past transcriptions from the new History window anytime.
Voice Commands
Say "new line", "comma", "exclamation mark", or "delete that" to format text hands-free. Add custom hotwords for industry-specific terms.
Redesigned Overlay
A beautiful rounded pill recording indicator with voice-reactive animated bars. Focus-safe: it never steals keyboard focus from your target app.
One-Click Installer
No more manual setup. The installer auto-detects Python 3.11, creates a virtual environment, and downloads all AI models with a clear progress bar.
Features

Power. Speed. Control.

Everything you need to dictate fast, privately, and with complete flexibility.

Instant Transcription
Near-zero latency via faster-whisper on-device models. Speaking is the most natural way humans communicate capture ideas at 130+ words per minute, up to 3x faster than typing.
100% Local and Private
No internet connection required for dictation. All AI inference runs on your hardware. Your voice data stays on your machine, always. No accounts, no telemetry, no tracking.
Works Everywhere
Runs quietly as a lightweight system tray app. Press the global hotkey and dictate directly into any text field emails, documents, chat apps, code editors, and more.
Global Hotkey
Ctrl + Shift + Space from any application. Fully customizable in Settings. Supports both Toggle mode and Hold-to-Talk mode for hands-free flexibility.
CPU-Only Ready
Runs on any modern CPU via int8 quantization. No GPU required. Just install and start speaking Typra handles the rest automatically without any configuration.
Free Forever
Open-source under the MIT License and 100% free to use. No SaaS fees, no usage limits, no vendor lock-in. You are free to use, share, and modify it. You own your tool.
Getting Started

From voice to text
in minutes

No setup headaches, no account creation. Download, install, and start dictating immediately.

01 / Install
Install Typra
Download Typra_v2_Setup.exe and run the installer. It automatically creates a virtual environment, and downloads all AI model files with a progress bar.
IMPORTANT: You need to have Python 3.11.9 installed. You can download it under the following link: Python 3.11.9
02 / Launch
Launch the App
Open Typra from your Desktop or Start Menu. It runs quietly in your system tray in the bottom-right corner. Right-click the icon to access all options and settings.
03 / Focus
Click Your Target
Click into the application where you want text to appear a document, email, chat window, or code editor. Then press the global hotkey to begin recording.
04 / Speak
Speak and Review
Speak naturally. Say "comma" or "new line" to format on the fly. Press the hotkey again to stop. Typra cleans up your text and types it automatically.
Default hotkey: Ctrl + Shift + Space customizable in Settings
Ready to start dictating?
Free. Offline. Private. No account required.
Download Typra v2.0.0 Older Versions
Windows 10 / 11 (64-bit) Approximately 3 GB free disk space · Python · Internet required during installation only
Requirement To use Typra you need to have Python version 3.11.9 installed. You can download it by clicking the following link: Python 3.11.9
Notice Typra is an early-release tool provided as-is. It may contain bugs or behave unexpectedly in some environments. Use at your own discretion. View source on GitHub.
Frequently Asked Questions

Got questions?
We have answers.

No. You only need an internet connection during the initial installation to download the AI models. Once installed, Typra works completely offline every time you use it. No data ever leaves your device.
Yes. Typra is open-source under the MIT License and 100% free forever. No subscriptions, no usage limits, no vendor lock-in. You are free to use, share, and modify it however you like.
Windows 10 or 11 (64-bit), approximately 3 GB of free disk space,Python version 3.11.9 installed, a working microphone, and an internet connection only during the first installation to download the AI models.
No. Typra requires no account, no registration, and no email address. Download the installer and the correctPython version, run it, and start dictating immediately. Completely anonymous by design.
Never. All audio and text processing happens 100% locally on your machine. There are no servers, no telemetry, no logging. No data of any kind ever leaves your device.
Your history is stored in a local SQLite database in the ~/.whisper_dictation/ folder on your machine. You can view, delete individual entries, or clear all history at any time from inside the app.
None whatsoever. Typra writes crash logs only to a local file at ~/.whisper_dictation/typra.log on your own machine. Nothing is transmitted anywhere. The code is open-source so you can verify this yourself.
No. Typra runs entirely on any modern CPU using int8 quantization. If you have a CUDA-capable GPU, Typra v2.0.0 will automatically use it for significantly faster transcription speeds.
If AI Cleanup is enabled, the local LLM takes 1 to 5 seconds on a CPU to analyze, correct grammar, and apply your selected tone before typing. For instant raw output, disable AI Cleanup in Settings.
The "base" model (145 MB) offers the best speed-to-accuracy ratio on CPU and is recommended for most users. Use "tiny" for maximum speed, or "small" and "medium" for higher accuracy when processing time is not a concern.
Say "new line", "comma", "period", "exclamation mark", or "delete that" while dictating and Typra interprets them as formatting commands rather than transcribing the words. Enable or disable voice commands in Settings.
Yes. Typra supports over 20 languages and you can switch between them in Settings. The active language guides Whisper toward correct transcription and activates the appropriate voice command vocabulary.
Yes. Open Settings and click the hotkey recorder field to capture any combination you prefer. You can also switch between Toggle mode (press once to start, again to stop) and Hold-to-Talk mode for maximum control.