Ollama Voice Mac Project Introduction
Overview
The Ollama Voice Mac project offers an offline voice assistant designed to operate on Mac systems. It utilizes two powerful tools: the Mistral 7b model through Ollama and Whisper speech recognition models. Originally inspired by the foundational work from another project, this version enhances its compatibility with Mac computers, incorporating several upgrades to improve functionality and user experience.
Key Features
- Offline Functionality: Ollama Voice Mac operates completely offline, ensuring privacy and independence from internet connectivity.
- Compatibility: Specifically designed for Mac users, offering a seamless integration into the Apple ecosystem.
- Customizable Language Support: Users have the option to configure the assistant to recognize and operate in various languages by modifying configuration settings.
Installation and Setup
Embarking on the journey to bring Ollama Voice Mac to life involves a straightforward installation process:
- Install Ollama: Begin by downloading Ollama on your Mac, which serves as the foundational platform for the assistant.
- Mistral 7b Model: Retrieve this AI model using the command
ollama pull mistral. - Whisper Model: Obtain a compatible OpenAI Whisper Model, like the base.en version, which provides robust speech recognition capabilities.
- Repo Cloning: Clone the project repository to your local machine.
- Model Placement: Place the downloaded Whisper model within a directory labeled
/whisperat the root of the cloned repository. - Essential Tools: Ensure the installation of Python and Pip, which are vital for running the assistant's software.
- Apple Silicon Compatibility: For those using Apple silicon, install Homebrew and execute
brew install portaudioto facilitate the PyAudio library. - Dependencies Installation: Run
pip install -r requirements.txtto secure necessary dependencies. - Launching the Assistant: Use
python assistant.pyto start the voice assistant.
Enhancing Voice Quality
To achieve a more refined auditory experience, users can upgrade the default voice settings:
- Navigate to System Settings, then Accessibility, followed by Spoken Content.
- Opt for Manage Voices and locate "Zoe (Premium)" among the English voices. Download and select this option as the system voice for a richer output quality.
Multilingual Capabilities
Users looking to expand the assistant’s language ability can modify the assistant.yaml configuration file. By adjusting the modelPath and incorporating a Whisper model in the desired language, the assistant can effectively communicate in languages other than English.
This project empowers users with a robust, private, and fully offline voice assistant experience tailored for Mac users, with flexibility for language customization and improved voice outputs.