SayPass – Voice is the New Password
Inspiration
Till now, we’ve commonly used patterns, numerical PINs, or fingerprint-based passwords for device and app authentication.
As an upgrade, SayPass introduces a smarter, more personalized way of authentication — your voice
Unlike many modern smartphones that recognize only voice commands , SayPass goes further by analyzing the unique vocal frequency, pitch, and tone of a person’s voice, something that's hard to imitate and unique to each individual.
What We Built
SayPass is a dual-verification voice authentication system that:
- Verifies who is speaking using ECAPA-TDNN speaker embeddings.
- Verifies what is spoken using Whisper transcription.
- Grants access only when both the voiceprint and phrase match.
Say your passphrase, and SayPass checks:
- Is it your unique voice frequency?
- Did you say the correct phrase?
Only when both match, access is granted ✓
Use Cases
- Cybersecurity defense – Adding a biometric voice layer to protect against unauthorized access
- Hands-free access in vehicles, workstations, or AR/VR setups
- Voice-based smart lock access
- Authentication for visually impaired users
- Forensic voice analysis – Matching suspect’s voice with crime scene audio evidence
What We Learned
- How to integrate voice biometrics using ECAPA-TDNN
- How to use Whisper for accurate transcription, even with accents
- The importance of combining voice tone + spoken content for high-security systems
Challenges Faced
- Background noise and silent voice trimming
- Fine-tuning thresholds to avoid false accept or false reject
The Future
We envision SayPass as a secure, scalable system that can be used in:
- Mobile & Web Authentication
- Smart Home & IoT Devices
- Personalized AI Assistants
Because in the future...
Your voice should be your password.
Built With
- ai
- openai
- resemblyzer
Log in or sign up for Devpost to join the conversation.