Alzheimer Voice
alzheimervoice is the world's first open-source cognitive voice fingerprint engine that detects early signs of Alzheimer's, Parkinson's, depression, Lewy Body Dementia, and Frontotemporal Dementia through voice analysis
Core trick
Extracts 107 biomarkers from natural speech, then runs a 35-rule differential over 11 conditions to score cognitive risk.
Description
alzheimervoice is the world's first open-source cognitive voice fingerprint engine that detects early signs of Alzheimer's, Parkinson's, depression, Lewy Body Dementia, and Frontotemporal Dementia through voice analysis alone. A 5-minute daily conversation is all it takes. The V5 engine extracts 107 biomarkers from natural speech across 11 cognitive domains. We sourced, transcribed, and ingested 80+ research papers to train the engine — making it uniquely evidence-grounded. Key innovations:
- Opus 4.6 dual-pass extraction with Extended Thinking — self-correction against 25 deterministic NLP anchors reduces variability by 40-50%
- 35-rule differential diagnosis across 11 conditions — the only tool performing linguistic differential diagnosis
- Topic-aware scoring eliminates false positives
- Weekly deep analysis with 32K Extended Thinking
The engine powers alzheimervoice.org for demo
README excerpt
# AlzheimerVoice ### 🏆 Cerebral Valley × Anthropic Hackathon — February 2026 > *"The voice remembers what the mind forgets."* **[alzheimervoice.org](https://alzheimervoice.org)** — Public site, live demo, science, open source **[Try the Live Demo →](https://alzheimervoice.org/demo)** — Record 30 seconds of speech, get an instant 11-domain cognitive voice analysis. **[CVF Engine (open source) →](https://github.com/remifrancois/cognitivevoicefingerprint)** — 107 indicators, 11 domains, 35 rules, 11 conditions, 84+ studies. --- 55 million people live with Alzheimer's worldwide — most diagnosed years too late, after irreversible neural damage. Research shows speech patterns change up to 7.6 years before clinical diagnosis (Eyigoz 2020). AlzheimerVoice detects early signs of Alzheimer's disease, Parkinson's, depression, Lewy Body Dementia, Frontotemporal Dementia, and 6 other conditions through voice analysis alone. The V5 "deep_voice" engine extracts a 107-indicator **Cognitive Voice Fingerprint (CVF)** from natural speech using multimodal analysis (acoustic + Claude transcription + NLP anchors + topic detection) and tracks cognitive drift over weeks and months. Built by **Rémi F.** during a 6-day hackathon with $500 in Claude Opus 4.6 API tokens. ## Live Demo The demo at **[alzheimervoice.org/demo](https://alzheimervoice.org/demo)** lets anyone: 1. **Record** 30-60 seconds of natural speech (French or English) 2. **Analyze** in real-time: GPU acoustic extraction → Cl