Deploy a Vultr Tortoise TTS instance. Then install the Tortoise TTS fast fork over top. https://github.com/152334H/tortoise-tts-fast
sudo make install_server
Edit /etc/athena/config.json with your key and server info.
Start the service with,
systemctl enable --now /opt/athena/athena-server.service
Its recommended to put this behind nginx.
This requires a few extra steps,
To start,
sudo make install_client
Go to, https://picovoice.ai/platform/porcupine/
Sign up for a free account. Now you need 3 wake words,
- "Athena Listen"
- "Athena Request"
- "Athena Quiet"
Train and download these wake words. Unzip the ppn files to these locations,
/opt/athena/wake/athena-listen.ppn/opt/athena/wake/athena-request.ppn/opt/athena/wake/athena-quiet.ppn
Now add your porcupine and server details to the config file,
/etc/athena/config.json
You are going to want to pick a voice, or leave it as freeman. Then you want to generate the pre-rendered voice responses.
sudo python3 /opt/athena/athena_generate.py
If you set up the server right, this should be working. If not, check the server!
Now start the assistant!
systemctl enable --now /opt/athena/athena.service
This should be working and you should be ready to go!
- "Athena Listen" is used to trigger commands preset in the config.
- "Athena Request" hits the server, uses OpenAI to get a response, and then TortoiseTTS to render the response
- "Athena Quiet" is to tell Athena to bail on a request or to stop talking during a long one
- "Never mind" tells Athena to stop waiting for a command
- "What are you" tells Athena to identify itself