LOCAL SPEECH PIPELINE · MODEL PROVIDER MAY BE REMOTE · WINDOWS AUTOMATION
privacy68
# Autonomous Local Voice Assistant & OS Automation
# Autonomous Local Voice Assistant & OS Automation
Illustrative command-flow demo. The current application sends most requests to the Agent Brain; this website does not execute commands.
[WEBSITE DEMO] This terminal is illustrative; it is not connected to the desktop app.
> [DEMO INPUT] "Privacy68, open Google Chrome"
> [CURRENT FLOW] Other commands go to the Agent Brain.
> [DEMO ONLY] No model request or desktop action is performed.
> _
Hardware-accelerated neural networks running directly on your Windows desktop.
Powered by faster-whisper (CTranslate2) with CUDA float16 GPU
acceleration and automatic int8 CPU fallback.
Speech processing is performed by the local speech pipeline. The configured Agent Brain model is cloud-tagged, so requests may be sent to a remote provider.
The Agent Brain requests Python code through LiteLLM using ollama/gemma4:31b-cloud, then executes returned code in a local Python subprocess.
32ms frame-level Voice Activity Detection ignores keyboard clicks, breathing, and background TV chatter seamlessly.
Full-screen transparent overlay with Windows WS_EX_TRANSPARENT integration so games & work are never blocked.
Manage built-in and trusted Python plugins in the desktop dashboard. Custom plugins are discovered from your user plugin folder.
Run directly from source or download the latest Windows installer.
irm https://raw.githubusercontent.com/PRN-6/PRIVACY68/main/install.ps1 | iex
git clone https://github.com/PRN-6/PRIVACY68.git && cd PRIVACY68
pip install -r requirements.txt
python app.py