Ollama Vox

A local macOS menubar voice assistant that records speech, transcribes with MLX Whisper, gets responses from Ollama, and speaks back using Kokoro TTS.

Features

• Menubar app for easy access
• Local speech-to-text with mlx-whisper
• Local text generation with Ollama
• Local text-to-speech with mlx-audio + Kokoro voices
• Status panel showing model info, rolling latency stats, and response history

Requirements

• macOS (Apple Silicon recommended)
• Python 3.12+
• Ollama installed and running locally

Installation

The easiest way to install is via pip or uv:

pip install ollama-vox
# or if using uv
uv tool install ollama-vox

If installing from source for development:

git clone https://github.com/maadhav-codes/ollama-vox.git
cd ollama-vox
uv sync

First-Time Setup

Download recommended STT + TTS models:

uv run ollama-vox --setup

Run

Start the menubar application:

uv run ollama-vox

Usage

1. Click the microphone icon in your macOS menubar.
2. Select Start Listening to speak.
3. Select Stop Listening when you are done. The app will process your speech and respond with audio.
4. Click Show Status to view latency, recent responses, and active models.

Configuration

Settings are managed in config.yaml.

• audio: Adjust Voice Activity Detection (VAD) and recording limits.
• stt.model: Path to the local Whisper model.
• ollama: Set the endpoint, model name, and temperature.
• tts: Configure the voice, speaking rate, and Kokoro model path.

Community

We welcome contributions and feedback from the community!

• Contributing: Please see our Contributing Guide for local setup instructions and our PR review workflow.
• Reporting Issues: Encountered a bug or have a feature idea? Please use our Issue Forms.
• Discussions: Have questions or want to share how you're using Ollama Vox? Join the conversation in GitHub Discussions.
• Security: If you find a security vulnerability, please follow the steps in our Security Policy to report it privately.