Voice authentication has long been plagued by the fundamental challenge of encoding complex, high-dimensional voice embeddings into blockchain-verifiable commitments. Traditional approaches struggle with the sheer volume of information contained in voice biometrics - a single voice sample can contain thousands of floating-point values representing pitch, timbre, cadence, and unique vocal characteristics. The ECAPA-TDNN (Emphasized Channel Attention, Propagation and Aggregation Time-Delay Neural Network) model generates these rich 192-dimensional embeddings, but existing blockchain solutions fail to efficiently verify voice similarity without compromising privacy or requiring massive computational overhead. The long-standing problem was that voice could not be decoded because of so much information and complex vector embeddings - we are solving that using ECAPA-TDNN, which captures the subtle nuances of human speech that make each voice unique, from the micro-variations in pronunciation to the distinctive rhythm patterns that define individual speaking styles.

Resona solves this by deploying a novel zero-knowledge proof system on Zircuit that can verify voice similarity while maintaining complete privacy. Our smart contracts store only cryptographic commitments of voice embeddings, while the ZK-proof system mathematically proves similarity without revealing the actual voice data. This breakthrough enables secure, decentralized voice authentication that scales to millions of users while preserving the mathematical precision of ECAPA-TDNN embeddings.

Zircuit's Layer 2 infrastructure provides the perfect foundation for Resona's voice authentication protocol by solving the fundamental challenge of on-chain vector embedding computation. Traditional blockchain systems could only handle simple cryptographic operations off-chain, but Zircuit's advanced rollup technology enables us to bring complex vector mathematics directly on-chain.

The Vector Embedding Revolution: Voice embeddings generated by ECAPA-TDNN are complex mathematical objects containing 192-dimensional vectors with floating-point precision. Before Zircuit, these computations were impossible on-chain due to:

• Computational Complexity: Vector similarity calculations require matrix operations that exceed traditional blockchain limits
• Precision Requirements: Voice recognition demands sub-millisecond accuracy that simple integer math cannot provide
• Scalability Constraints: Processing thousands of voice embeddings simultaneously was computationally prohibitive

Zircuit's ZK-rollup architecture transforms this limitation by enabling:

• On-chain Vector Operations: Direct computation of cosine similarity, Euclidean distances, and complex mathematical functions
• Real-time Verification: Sub-second voice similarity verification with mathematical precision
• Scalable Processing: Handling millions of voice embeddings without compromising performance

The network's advanced rollup technology handles the complex ZK-proof verification at scale, while its EVM compatibility ensures seamless integration with existing Web3 infrastructure. By deploying on Zircuit, Resona achieves sub-second voice verification with minimal gas costs, making voice-based authentication accessible to mainstream applications for the first time.

Deployed block explorer link: https://explorer.garfield-testnet.zircuit.com/address/0xAC3a3123770346d8d4c186d892d81b4522b1D512

Resona leverages The Graph's powerful indexing capabilities to create an intelligent, adaptive voice authentication system with efficient data storage and fetching. Our subgraph implements sophisticated data optimization techniques that enable lightning-fast queries while maintaining comprehensive voice analytics. By leveraging The Graph's advanced indexing algorithms, we achieve sub-second response times for complex queries involving thousands of voice registrations and authentication attempts.

The system employs intelligent data partitioning and lazy loading strategies to minimize storage overhead while maximizing query performance. Our subgraph monitors suspicious activities, tracks voice pattern changes over time, and implements dynamic threshold algorithms that automatically adjust security parameters based on user behavior. The system calculates optimal authentication thresholds for each wallet, increasing security measures when detecting failed attempts or unusual voice variations. This AI-driven approach transforms raw blockchain data into actionable security insights, enabling proactive threat detection and personalized authentication experiences that adapt to each user's unique voice characteristics and usage patterns.

Key Data Efficiency Features:

• Smart Indexing: Optimized database schemas for voice-specific queries
• Compressed Storage: Efficient encoding of high-dimensional voice data
• Real-time Updates: Live indexing of blockchain events with minimal latency
• Scalable Architecture: Handles millions of voice registrations without performance degradation

Query our live subgraph to see the intelligent voice analytics in action:

query {
  voice(id: "0xb5f94326908cf0110171f98fc87f75ff0573f321216c62709aaa4dc63a426fb9") {
    id
    owner
    walrusUri
    timestamp
    thresholdAnalytics {
      id
      totalAttempts
      successfulAttempts
      failedAttempts
      currentOptimalThreshold
      averageSuccessfulSimilarity
      averageFailedSimilarity
      securityScore
      lastCalculated
      recommendedThreshold
    }
    authAttempts {
      id
      success
      similarity
      threshold
      timestamp
      isAboveOptimalThreshold
      riskScore
    }
  }
}

This query demonstrates our advanced analytics engine that automatically calculates optimal security thresholds, tracks authentication patterns, and provides real-time risk assessments for each voice identity.

Example Response:

{
  "data": {
    "voice": {
      "id": "0xb5f94326908cf0110171f98fc87f75ff0573f321216c62709aaa4dc63a426fb9",
      "owner": "0x19e95b026731974b7c1fed9eb3c3113fbdd80464",
      "walrusUri": "NEnvIzmJWTLA7j66wvgEb5I_tG_CzzpJOwpJ3-mkGz0",
      "timestamp": "1755397676",
      "thresholdAnalytics": {
        "id": "0xb5f94326908cf0110171f98fc87f75ff0573f321216c62709aaa4dc63a426fb9",
        "totalAttempts": "1",
        "successfulAttempts": "1",
        "failedAttempts": "0",
        "currentOptimalThreshold": "8258",
        "averageSuccessfulSimilarity": "8758",
        "averageFailedSimilarity": "0",
        "securityScore": "3000",
        "lastCalculated": "1755401331",
        "recommendedThreshold": "8758"
      },
      "authAttempts": [
        {
          "id": "1755401325958",
          "success": true,
          "similarity": "8758",
          "threshold": "7500",
          "timestamp": "1755401325",
          "isAboveOptimalThreshold": true,
          "riskScore": "2000"
        }
      ]
    }
  }
}

Resona revolutionizes Web3 authentication by integrating Dynamic's embedded wallet infrastructure with advanced AI-powered voice recognition to enhance the wallet experience beyond traditional authentication methods. Our system transforms Dynamic's seamless wallet management into an intelligent, voice-first experience that eliminates the friction of traditional Web3 onboarding.

Enhanced Wallet Experience Features:

• Voice-First Authentication: Users can access their Dynamic wallets using just their voice, eliminating the need for seed phrases or complex key management
• Intelligent Security: AI continuously learns from voice patterns, automatically adjusting security thresholds based on user behavior and risk assessment
• Seamless Integration: Dynamic's embedded wallet infrastructure handles complex operations while Resona provides the voice authentication layer
• Personalized Security: Each user gets custom security parameters that adapt to their unique voice characteristics and usage patterns

The AI-Powered Wallet Revolution: Our system uses Dynamic for seamless wallet management while leveraging ECAPA-TDNN AI models to create a truly intelligent authentication experience. The AI component continuously learns from user voice patterns, automatically adjusting security thresholds and detecting anomalies in real-time. This creates a crypto-native experience where users can authenticate using just their voice, with Dynamic handling the complex wallet operations behind the scenes.

Beyond Traditional Authentication:

• Zero-Knowledge Voice Verification: Prove voice ownership without revealing actual voice data
• Adaptive Security: Security thresholds that automatically adjust based on successful/failed attempts
• Risk Assessment: Real-time calculation of authentication risk scores and security recommendations
• Behavioral Learning: AI that understands your voice patterns and adapts security accordingly

The AI system acts as an autonomous agent that monitors authentication attempts, learns from user behavior, and provides personalized security recommendations, making Resona a natural extension of Dynamic's vision for intelligent, user-friendly Web3 experiences. This integration creates a new paradigm where voice becomes the primary key to your digital identity, seamlessly integrated with Dynamic's powerful wallet infrastructure.

• Node.js (v18 or higher)
• Python 3.8+
• Git
• MetaMask or other Web3 wallet
• Zircuit testnet RPC endpoint

1. Clone the repository

   git clone https://github.com/your-username/resona.git
   cd resona

2. Install backend dependencies

   cd backend
   npm install

3. Install frontend dependencies

   cd ../frontend
   npm install

4. Install smart contract dependencies

   cd ../contracts
   npm install

5. Install voice server dependencies

   cd ../voice-server
   python -m venv venv
   source venv/bin/activate  # On Windows: venv\Scripts\activate
   pip install -r requirements.txt

6. Install subgraph dependencies

   cd ../subgraph
   npm install

The voice server handles voice embedding generation and similarity calculations:

cd voice-server
source venv/bin/activate  # On Windows: venv\Scripts\activate
python main.py

The server will start on http://localhost:8000 and provide endpoints for:

• Voice embedding generation
• Similarity calculations
• Model inference

Deploy the voice registry contracts to Zircuit testnet:

cd contracts
npx hardhat compile
npx hardhat run scripts/deploy.js --network zircuit-testnet

Important: Update the contract addresses in your configuration files after deployment.

Deploy the subgraph to The Graph's hosted service:

cd subgraph
npm run codegen
npm run build
npm run deploy

The backend provides the API layer for the frontend:

cd backend
npm run dev

The server will start on http://localhost:3001 with endpoints for:

• Voice authentication
• ZK-proof generation
• Blockchain interactions

Launch the React frontend application:

cd frontend
npm run dev

The application will be available at http://localhost:5173

Our subgraph is deployed on The Graph's hosted service and can be accessed at:

• Subgraph URL: https://thegraph.com/hosted-service/your-subgraph-name
• GraphQL Endpoint: https://api.thegraph.com/subgraphs/name/your-subgraph-name

For development and testing, you can also use Subgraph Studio:

• URL: https://studio.thegraph.com/
• Local Development: Run npm run dev in the subgraph directory

Test the subgraph with these example queries:

Get all voice registrations:

query {
  voices {
    id
    owner
    walrusUri
    timestamp
  }
}

Get authentication attempts for a specific voice:

query {
  authAttempts(where: { voice: "0x..." }) {
    id
    success
    similarity
    threshold
    timestamp
  }
}

Get threshold analytics:

query {
  thresholdAnalytics {
    id
    totalAttempts
    successfulAttempts
    currentOptimalThreshold
    securityScore
  }
}

cd contracts
npx hardhat test

cd voice-server
python -m pytest tests/

cd frontend
npm run test

cd backend
npm run test

Create .env files in each directory with the following variables:

Backend (.env):

ZIRCUIT_RPC_URL=your_zircuit_rpc_url
PRIVATE_KEY=your_private_key
CONTRACT_ADDRESS=deployed_contract_address

Frontend (.env):

VITE_BACKEND_URL=http://localhost:3001
VITE_GRAPH_URL=https://api.thegraph.com/subgraphs/name/your-subgraph-name

Voice Server (.env):

MODEL_PATH=./pretrained_models/spkrec-ecapa-voxceleb
PORT=8000

Update hardhat.config.js in the contracts directory with your network settings:

module.exports = {
  networks: {
    zircuitTestnet: {
      url: process.env.ZIRCUIT_RPC_URL,
      accounts: [process.env.PRIVATE_KEY],
      chainId: 80001, // Update with actual Zircuit chain ID
    }
  }
}

1. Connect your wallet to the frontend
2. Navigate to "Register Voice"
3. Record your voice sample (minimum 3 seconds)
4. Submit the registration transaction
5. Wait for confirmation on Zircuit

1. Navigate to "Authenticate Voice"
2. Record your voice sample
3. The system will generate a ZK-proof
4. Submit the authentication transaction
5. View results and analytics

1. Go to "Voice Analytics"
2. See your authentication history
3. View security scores and recommendations
4. Monitor threshold adjustments