Weebo

A real-time speech-to-speech chatbot powered by Whisper Small, Llama 3.2, and Kokoro-82M.

Works on Apple Silicon.

Learn more here.

Dependancies

First get uv

1
  curl -LsSf https://astral.sh/uv/install.sh | sh

Make sure you have rust

1
  curl --proto '=https' --tlsv1.2 -sSf https://sh.rustup.rs | sh

Make sure you have ollama

1
  brew install ollama

Make sure you have llama3.2

1
  ollama pull llama3.2

Make sure you have espeak-ng

1
2
  brew install espeak-ng
  export ESPEAK_DATA_PATH=/opt/homebrew/share/espeak-ng-data

Get the code

1
2
  git clone https://github.com/amanvirparhar/weebo
  cd weebo

Pull down kokoro

1
  wget https://github.com/thewh1teagle/kokoro-onnx/releases/download/model-files/kokoro-v0_19.onnx

Fire it up

1
  uv run --python 3.12 --with-requirements requirements.txt main.py

It will download the whisper weights when it first starts.

Doesn't understand what you say sometimes and you can't interrupt it, but super cool that it all runs locally!