Attempt at creating a virtual assistant that can run locally and that could be used for learning various things, mostly languages as a first focus.
Currently the system uses:
- A large language model:
Llama-3 8B
- A text-to-speech model:
xtts_v2
- A speech-recognition model:
whisper large-v3
First, install all the dependencies by running the following commands:
conda create python=3.10 --name="virtual-teacher" -y
conda activate virtual-teacher
pip install pip-tools
pip-compile requirements.in
pip install -r requirements.txt
Then, you can check the available languages by running:
python main.py --list-languages
Finally, choose a language (e.g. es
for Spanish) and run the following command:
# If you do not want to use speech recognition
python main.py --language=es
# If you want to use speech recognition
python main.py --language=es --use-speech-recognition
Note: If you ever get an error when recording your voice, just try again. I've been experiencing some issues as well but it usually works the second time.