🎤 Scriba Ro

Offline voice dictation assistant for Windows.
Press a key, speak, and the text appears automatically in any application.

---

What it does

• Romanian dictation — uses upb-nlp/ro-fast-conformer (NeMo, WER 3.01%)
• English dictation — uses openai/whisper-large-v3-turbo (CTranslate2)
• 100% offline — no audio ever leaves your machine
• Auto-inject — transcribed text is pasted directly into the active window via clipboard
• Intelligent post-processing — automatic punctuation and capitalization restoration for Romanian (using punctuators library) and smart spacing injection
• System tray — runs quietly in the background, no taskbar clutter
• Floating overlay — small visual indicator when recording or transcribing
• Quick toggle — switch language with F8 or via the Dashboard button

Requirements

• Windows 10/11
• Python 3.10+
• NVIDIA GPU with CUDA (recommended, CPU also works)
• ~13 GB free disk space (models + virtual environment)

Installation

# 1. Clone the repository
git clone https://github.com/gal101/scriba-ro.git
cd scriba-ro

# 2. Run the setup wizard (downloads models and configures the environment)
run.bat

On first run, launcher.py automatically checks and installs dependencies, while setup.py downloads the Whisper model.
The NeMo model for Romanian is downloaded separately (see AGENT.md for details).

Usage

1. Run run.bat
2. A red icon appears in the system tray and the Dashboard window opens
3. Hold down the configured key (default: Right Control) and speak
4. Release the key — the transcribed text appears in the active window
5. Switch language with F8 or the green/blue button in the Dashboard

License

Personal project. AI models are subject to their own licenses (Apache 2.0 / CC-BY).