ollama-voice-mac

A completely offline voice assistant using Mistral 7b via Ollama and Whisper speech recognition models. This builds on the excellent work of maudoin by adding Mac compatibility with various improvements.

https://github.com/apeatling/ollama-voice-mac/assets/1464705/996abeb7-7e99-451b-8d3b-feb3fecbb82e

Installing and running

1. Install Ollama on your Mac.
2. Download the Mistral 7b model using the ollama pull mistral command.
3. Download an OpenAI Whisper Model (base.en works fine).
4. Clone this repo somewhere.
5. Place the Whisper model in a /whisper directory in the repo root folder.
6. Make sure you have Python and Pip installed.
7. For Apple silicon support of the PyAudio library you'll need to install Homebrew and run brew install portaudio.
8. Run pip install -r requirements.txt to install.
9. Run python assistant.py to start the assistant.

Improving the voice

You can improve the quality of the voice by downloading a higher quality version. These instructions work on MacOS 14 Sonoma:

1. In System Settings select Accessibility > Spoken Content
2. Select System Voice and Manage Voices...
3. For English find "Zoe (Premium)" and download it.
4. Select Zoe (Premium) as your System voice.

Other languages

You can set up support for other languages by editing assistant.yaml. Be sure to download a different Whisper model in your language and change the default modelPath.