A real-time voice AI assistant that runs entirely on your machine. No cloud. No latency. No compromises.
Built with the technologies that matter.
Whisper Turbo MLX delivers blazing fast speech-to-text on Apple Silicon. No waiting, no buffering.
Powered by Llama 3.1 running locally via llama.cpp. Your conversations never leave your device.
No API keys. No cloud services. No internet required. Pure, standalone AI.
MLX framework ensures maximum performance on M1, M2, M3, and M4 chips.
Three steps. One seamless experience.
Just talk naturally. TARS captures your voice in real-time using your microphone.
Whisper transcribes your speech instantly. The LLM comprehends context and intent.
Natural speech synthesis delivers the response. A fluid conversation, powered by AI.
A carefully crafted architecture for voice AI.
Clone, install, run. It's that simple.
# Clone the repository
$ git clone https://github.com/yourusername/tars.git
$ cd tars
# Install dependencies
$ pip install -r requirements.txt
# Download a model (example: Llama 3.1 8B)
$ mkdir -p models && cd models
$ wget https://huggingface.co/...
# Run TARS
$ python index.py