Tech for Good: Giving a Voice to the Voiceless

Your voice is an essential part of your identity. It can be incredibly frustrating and isolating if you can’t express yourself and your thoughts. Fortunately, there are technologies to help people who have lost their voices due to cancer, ALS, and other conditions. US AI MVP Charles EllwoodHe has used Microsoft’s app to help several people regain their voices. Custom neural voice (CNV) and AI. The app converts text to speech with a custom natural-sounding synthesized voice.

Former radio DJ and manager Chris Martin lost his ability to speak due to throat cancer and a surgical removal of his larynx. Because Chris was a DJ, Charles had recordings of his voice that he could use in an app that would allow Chris to speak again using AI.

Radio DJ and manager Chris Martin (left) interviews AI MVP Charles Ellwood (right)

Charles created a “bank” of Chris’ voices so he could input the phrases he wanted to say on his computer or phone. This text-to-speech service allows Chris to share humor, life stories, and have deep conversations with his wife and family. He can even watch live sports broadcasts, something he enjoyed doing when he was a DJ. Chris was able to interview Charles on radio using his app. Chris’s wife told Charles, “Charlie, you brought him home.”

Q&A with Charles about combining his passion for the greater good with his technical expertise to help Chris and others speak with their own voice:

MVP Program: Please share some background on your recent projects using AI and Microsoft Neural Voice.
Charles: I help people with cerebral palsy.Amyotrophic lateral sclerosis (ALS), throat cancer, congenital diseases, and non-verbal disorders on the spectrum. These are terrible diseases that block communication, even though the brain is functioning and alive. I want to give those affected a voice and a chance to participate in our society again. Custom Neural Voice, Azure AI Custom Vision, and Azure technologies make it possible.

MVP Program: How do you think this technology will develop and how can it help people with speech impediments in the future?

Charles: I think for people who need mobility, smaller samples will be developed and run on the edge, allowing people to move around freely, with better privacy and security, and using less energy than large language models. I’m watching. Pie-3 Development and GPT4o Mini Excited! Fine-tuning, Retrieval Augmented Generation (RAG), and Small Language Models (SLM) are becoming more integrated and easier to connect. Pricing and models are improving to accommodate these cases. What’s interesting is that all of this is very new technology and there has already been a lot of progress in just two years.

MVP Program: How do you ensure that other people’s voices are used ethically? And what steps do you take to protect users’ privacy?

Charles: I leave ownership and decision-making of the voice to the original creator of the voice. I will discuss ownership with them in the future. Do they leave the voice in their will and give ownership to their spouse or children? I explain to them that we use Azure to host the voice, create a login to Azure with MFA (multi-factor authentication), and that they are the only people in the world who can access the voice. I also hear from many people who have lost a parent and want to hear their parent’s voice again. Sometimes to complete the grieving process, to help with closure, sometimes to help their children hear their parent’s voice. All of these findings help us identify where ethical issues may arise and educate people about what they need to know, and maybe this can change some of the rules and social structures surrounding ethics.

For more information on Charles Elwood’s project, see: