How harshitdevin/voiceguard works
VoiceGuard is a browser-based voice-authenticity detector with an optional Flask/librosa analysis backend. The main workflow takes uploaded audio or a live microphone stream, analyzes it across six signal features, then produces an authenticity score, verdict, and report; live capture also displays a waveform. The graph preserves the six documented analysis stages and the backend alternative. The UI and its wiring are unsampled, so their connections are documented or marked unverified rather than inferred from filenames.
Main entry points: server.py Analysis endpoint · analyzer.js Browser analysis engine · index.html Web interface · realtime.js Live capture
Components 22
Audio inputs
- Web interface
index.html - Live capture
realtime.js
Browser analysis
- Browser analysis engine
analyzer.js - Benford analysis
benford.js - Spectral smoothness
analyzer.js - Pitch jitter shimmer
analyzer.js - Noise floor
analyzer.js - Formant transitions
analyzer.js - Rhythm pauses
analyzer.js
Results
- Score and verdict
analyzer.js - Analysis report
analyzer.js
Optional Python backend
- Analysis endpoint
server.py - Benford analysis
server.py - Spectral smoothness
server.py - Pitch jitter shimmer
server.py - Noise floor
server.py - Formant analysis
server.py - Rhythm pauses
server.py - Score and verdict
server.py - Analysis report
server.py
Other components
- VoiceGuard user
- Microphone
Connections 23
- VoiceGuard user → Web interface: uses
README.md - VoiceGuard user → Microphone: speaks into
README.md - Microphone → Live capture: provides audio
realtime.js - Web interface → Live capture: starts capture
- Live capture → Browser analysis engine: analyzes audio
realtime.js - Web interface → Browser analysis engine: analyzes upload
- Browser analysis engine → Benford analysis: uses
analyzer.js - Browser analysis engine → Spectral smoothness: runs
README.md - Browser analysis engine → Pitch jitter shimmer: runs
README.md - Browser analysis engine → Noise floor: runs
README.md - Browser analysis engine → Formant transitions: runs
README.md - Browser analysis engine → Rhythm pauses: runs
README.md - Browser analysis engine → Score and verdict: combines scores
README.md - Score and verdict → Analysis report: produces
README.md - Web interface → Analysis endpoint: submits audio
- Analysis endpoint → Benford analysis: calls
server.py - Analysis endpoint → Spectral smoothness: calls
server.py - Analysis endpoint → Pitch jitter shimmer: calls
server.py - Analysis endpoint → Noise floor: calls
server.py - Analysis endpoint → Formant analysis: calls
server.py - Analysis endpoint → Rhythm pauses: calls
server.py - Analysis endpoint → Score and verdict: computes
server.py - Analysis endpoint → Analysis report: builds
server.py