Speech-to-text and text-to-speech technologies make the groundwork for voice-based user interactions, which are becoming more and more prevalent in consumer technology. Yet current STT and TTS solutions are proprietary, costly, and all require sending private user data to third-parties. Building state-of-the-art speech models requires incredible amounts of data and powerful hardware to train on, slowing down innovation. At Mozilla, we aim to change this by creating high quality, privacy-respecting, open source speech technology. In this talk, I will summarize how we've built our STT and TTS projects, show some cool things people have built with it, talk about our future plans and describe how to get involved.
Reuben Morais is a Senior Research Engineer working on the Machine Learning group at Mozilla. He started programming by reverse engineering and modifying games in high school. Acquiring a likeness for low-level hackery, he joined Mozilla to work as a systems programmer on the Firefox OS project. He then moved into the Machine Learning group to work on Mozilla's voice technology.