VoiceVault

VoiceVault's mission is to ensure no patient loses their identity to a disease that takes their voice. Over 200,000 Americans experience significant speech loss annually from laryngeal cancer, ALS, stroke, and neurological disease. AI voice cloning can now recreate a person's voice indistinguishably from one hour of audio, but no clinical product delivers this in real time. VoiceVault records patients before they lose their voice, builds a personalized voice clone, and gives them a text-to-speech app fast enough for real conversation and simple enough for daily use. Losing your voice should never mean losing yourself.

I don't remember my father's voice. He battled throat cancer since I was five, and surgery left him with an electrolarynx. It gave him a robotic voice that scared my mom and caused him physical pain every time he spoke. I ordered for him at restaurants, took his phone calls, and received texts instead of words even when we were in the same room. His electrolarynx would clog frequently, sometimes every hour, leaving him unable to speak at all unless he spent 15-20 minutes cleaning the hole in his throat. Fifteen years of watching my father suffer with no alternative left me building VoiceVault. I'm a computer science sophomore at Texas A&M, a returning AWS software engineering intern, and have attended selective programs at Y Combinator and Jane Street. My advisor Junaid Husain, a Harvard-trained neuroscientist and CEO of HealthConnect Texas, is connecting VoiceVault to clinics in the Texas Medical Center for our first pilot patients. Over 200,000 Americans experience significant speech loss annually from laryngeal cancer, ALS, stroke, and progressive neurological disease. VoiceVault's beachhead is laryngeal cancer: around 13,000 diagnoses per year, patients who permanently lose their larynx and need a communication tool for the rest of their lives. The immediate expansion is ALS: 33,000 Americans, every one of whom will eventually lose speech. Tobii Dynavox's open SDK gives VoiceVault a direct path to embed into the eye-tracking devices ALS patients already use. Beyond that, 120,000 stroke survivors per year develop dysarthria, and over one million Americans with Parkinson's experience progressive speech deterioration.VoiceVault solves voice loss end to end. Patients complete a 120-prompt recording protocol capturing their full natural speech range. A working clone is available after the first session and improves as recordings are added. The text-to-speech app targets sub-300 millisecond latency, with synthesis triggering on every spacebar press, streaming output while the patient types the next word. Our recording protocol captures significantly more speech data than any competing platform, producing a higher-fidelity clone. ModelTalker requires 1,600 phrases over seven hours and users report hating their banked voice. Apple's Personal Voice is too slow for daily conversation. No existing product combines clinical-grade voice banking with real-time synthesis fast enough for natural conversation. The platform is HIPAA compliant from day one. VoiceVault is free to the patient. We are pursuing PDAC verification for HCPCS code E2511, which reimburses $3,200 per patient. Our platform is FDA Class II under Product Code ILQ, which is 510(k) exempt, bypassing the multi-year premarket process while maintaining FDA compliance. Cloud cost per voice model is under $20, producing gross margins above 95%. Laryngectomy patients never recover their voice and churn occurs only at end of life. Year 1 targets 30 patients for $96,000 in revenue. Year 3 targets 700 patients across multiple clinical sites for $2.24 million. By Year 5, expansion into ALS through Tobii Dynavox integration brings revenue past $6 million. By Year 7, with stroke and Parkinson's populations entering the pipeline, VoiceVault targets over $12 million in annual revenue.