Project Description

I have some experience with VUI's and am interested in building my own from scratch using open source libraries.

This project would be aimed at learning more about the open-source offerings for creating a voice agent without relying on third-party paid API's, and in the process creating a simple voice agent embedded within an RPI4.

Hopefully this project can provide some simple utilities, such as asking for the weather in a particular city, what the time is, or even what my calendar looks like for the day. Creating a perfectly programmed voice agent is not the goal, but getting something working with open source tools would be really cool.

Goal for this Hackweek

Learn more about how speech is modeled and processed, and what recent open source offerings are available for creating free voice interfaces. Hopefully, have at least a semi-functional voice agent that can detect a wake-word ('geeko', if possible) and read back the weather in particular cities, as well as tell the time.

Resources

voice-to-json: https://github.com/synesthesiam/voice2json, for voice processing and intent / slot detection

mycroft-percise: https://github.com/MycroftAI/mycroft-precise, for training / using prebuilt wake-words.

Mimic text-to-speech: https://github.com/MycroftAI/mimic1 for light-weight TTS that can run on an RPI

Anyone is open to join, but this project will require some hardware (RPI4, microphone, speaker) and some NLP knowledge (or a willingness to learn )