Since May 2026, AImpower.org has partnered with the Speech Accessibility Project to advance voice recognition technology that works for people who stutter and people who are deaf or hard of hearing.
Trained and optimized for fluent speech, most speech AI systems today don’t perform well for people who stutter. Through this groundbreaking initiative with academia and industry partners, we aim to change that. We have worked with over a hundred people who stutter and record over a hundred hours of authentic, diverse stuttered speech, to help researchers and developers build more inclusive speech AI systems – systems that recognize, respect, and respond to stuttered speech.
“This project has already shown improvements in voice recognition technology for people with diverse voices,” said Mark Hasegawa-Johnson, professor of electrical and computer engineering at the University of Illinois Urbana-Champaign and leader of the Speech Accessibility Project. “We’re looking forward to working with members of the stuttering community to make sure they’re able to use their voices with technology, as well.”
Why We’re Joining This Effort
At AImpower.org, we are committed to bringing tangible benefits to the communities we serve.
In our research with the stuttering community, we’ve consistently heard frustrations with everyday speech AI products, along with a strong desire to interact with these technologies as effectively as fluent speakers do. This partnership is our response to that call, and a reflection of the disability movement’s motto: “Nothing About Us Without Us”.
Our Role in the Project
We’re excited to serve as the intermediary between the stuttering community and the researchers and developers building next-generation speech AI. While actively outreaching and communicating the goals of this initiative to the community, we see our primary role as an advocate for the stuttering community, representing your lived experiences, aspirations, and perspectives in shaping the future of speech AI technologies.
Our Goals
- Representation: Ensure stuttered speech is meaningfully and authentically represented in the dataset, capturing the rich diversity and variability of stuttering as it occurs in real life.
- Empowerment: Design a data collection experience where every participant feels heard, respected, and valued—not just as a data point, but as a contributor to a public and communal good.
- Connection & Awareness: Create space for meaningful dialogues – between stutterers and allies—about identity, challenges, and joy of stuttering. In our past work, we’ve seen lasting bonds and a deep sense of belonging emerge from these conversations. Together, we will build not only technology, but also friendships and public awareness.
📣 To the stuttering community: Our voice matters. Our voice belongs. Our voice can change the future of AI.
We will continue data collection into 2027 and ensure that we have a good representation of stuttering to shape the speech systems of tomorrow. Together, we can make AI learns that our stutter is not a defect—it’s a different way of speaking.
Follow the progress on this work on AImpower.org and get involved through this form.

