Uncover The Invisible.
TruthTone++ is a state-of-the-art deepfake detection platform designed to rebuild trust in the digital age. Beyond simple binary classification, TruthTone++ provides detailed forensic insights using 3D spectral visualization, timeline confidence mapping, and blockchain-backed authentication.
Developed in a 36-hour sprint at CxC (Compute by Create) 2026, the premier data science hackathon at the University of Waterloo, TruthTone++ beat out over 350 teams to win Best Audio AI Hack.
Click above to watch the demo on YouTube
- Multi-Layered AI Detection: Custom ML models trained on the ASVspoof 2019 dataset extract MFCCs and spectral features to detect generative artifacts with 99.8% accuracy.
- 3D Audio Fingerprinting: Revolutionary interactive 3D meshes (built with Three.js) that reveal the geometric regularities of AI-generated sound vs. the organic chaos of human speech.
- Timeline Analysis: A second-by-second confidence heatmap highlighting precisely where an audio clip might have been tampered with.
- Solana-Backed Verification: Immutable certificate minting on the Solana blockchain (via Metaplex) to create permanent, tamper-proof records of verified organic audio.
- Explanatory AI: Integration with Google Gemini to provide human-readable forensic reports, translating complex signal analysis into plain English.
- Deepfake Comparison: Direct integration with ElevenLabs for generating and comparing "visual diffs" between original and synthetic audio.
- Framework: Python / FastAPI
- Signal Processing: Librosa for spectral analysis and feature extraction.
- AI Models: Custom ML classifier + Google Gemini API for natural language analysis.
- Generation: ElevenLabs API for baseline deepfake creation.
- Framework: Next.js (TypeScript) + Tailwind CSS.
- Visualization: Three.js & React Three Fiber for WebGL-based 3D spectral rendering.
- User Experience: Real-time audio recording and drag-and-drop dashboard.
- Network: Solana
- Protocol: Metaplex for NFT-based authentication certificates.
- Security: SHA-256 file hashing for immutable on-chain records.
- Node.js: 18.x or higher
- Python: 3.10 or higher
- API Keys: Google Gemini & ElevenLabs (Optional for comparison feature)
-
Clone the project
git clone https://github.com/beanscrack/Truth-Tone.git cd Truth-Tone -
Configure Environment Create a
.envfile in the root directory:cp .env.example .env # Add your GEMINI_API_KEY to the .env file -
Launch the Platform TruthTone++ includes an automated setup script that handles the venv, dependencies, and starting both servers:
chmod +x setup_and_run.sh ./setup_and_run.sh
The backend will run on localhost:8000 and the frontend on localhost:3000.
- Live Call Detection: Real-time analysis for VoIP and phone calls.
- Browser Extension: Instant verification for social media and news sites.
- Multi-Modal Support: Expanding detection to video deepfakes.
- Enterprise API: Plug-and-play forensic auditing for legal and journalistic firms.
TruthTone++ provides a probabilistic assessment of audio authenticity. While highly accurate, no detection system is 100% infallible. Results should be treated as forensic indicators rather than absolute proof.