Project
Communication powered by open source research.
Euphonia
Project Euphonia
Communication powered by open source research.
Project Euphonia is a Google Research initiative working to make speech recognition software more accessible.
Conversations are a daily occurrence, but talking can be a major challenge for millions of people around the world with atypical speech.
How it works
Contribute
Learn more about our research partnerships with the Speech Accessibility Project and how you can help curate and develop open-source datasets.
Build
Leverage our suite of open-source materials to build and research more accessible speech recognition technologies.
Enhance
Work to integrate this research into app updates and other accessibility initiatives for people with non-standard speech.
Euphonia in action
Our work with Tim Shaw, a former NFL player living with ALS
How a former NFL player is using AI to get his voice back
Customized speech recognition for people with non-standard speech
many who struggle to communicate their thoughts to the world.”
Google Assistant to gain my independence…This program means so much to me and other people like me.”
help not only myself but also a larger group of people with
communication challenges that are often left out.”
Before you get started
How can I access open source resources to build speech recognition technologies?
How can I access open source resources to build speech recognition technologies?
You can access and contribute to Project Euphonia via GitHub today. Project Euphonia is a research initiative that allows developers in the ecosystem to build their own Automatic Speech Recognition (ASR) models for the purpose of research. We hope this will enable more people to develop tailored solutions for additional languages and cultural contexts.
How can I contribute to Google’s research?
How can I contribute to Google’s research?
To contribute to a diverse dataset of non-standard speech, we recommend reaching out to the Speech Accessibility Project. This initiative, led by the University of Illinois, Urbana-Champaign and supported by companies including Google, aims to improve speech recognition across the industry by making diverse data available to researchers and developers. The program is currently available in English.
How can I access non-standard speech data?
How can I access non-standard speech data?
While Project Euphonia’s dataset is not available to outside teams or companies, we are committed to supporting the ecosystem of speech recognition technologies through our Project Euphonia GitHub page. You can build your own audio tool to collect speech data, and train and transcribe functionalities for personalized Speech Models.