Skip to main content

​
Audio Processing

SAMMY Three includes advanced audio processing capabilities to ensure high-quality voice interactions in any environment.

​
Overview

The audio processing system provides multiple layers of enhancement to deliver crystal-clear voice communication.

Noise Suppression

Remove background noise with browser-native or AI-powered suppression

Noise Gate

Software-based filtering to eliminate ambient sounds

Environment Presets

Pre-configured settings optimized for different environments

​
Configuration

Enable audio processing in your SAMMY Three configuration. 
const config = {
  audioConfig: {
    noiseSuppression: {
      enabled: true,
      enhancementLevel: 'medium',
    },
    noiseGate: {
      enabled: true,
      threshold: 0.04,
    },
    environmentPreset: 'office',
  },
};

​
Noise Suppression

Two powerful noise suppression modes are available to match your needs.
  • Browser Native
  • Koala AI

​
Default Browser Suppression

Uses the browser’s built-in echo cancellation and noise suppression capabilities.
const config = {
  audioConfig: {
    noiseSuppression: {
      enabled: true,
      enhancementLevel: 'light', // 'light', 'medium', 'aggressive'
    },
  },
};
Pros:
  • No additional configuration required
  • Works on all modern browsers
  • Zero latency
  • Free to use
Cons:
  • Limited effectiveness in very noisy environments
  • Quality varies by browser and device

​
Enhancement Levels

Choose the right suppression level for your environment:

Light

Best for: Quiet offices, home officesMinimal processing that preserves natural voice quality while removing light background noise.
enhancementLevel: 'light'

Medium

Best for: Open offices, cafesBalanced approach that removes moderate noise while maintaining voice clarity.
enhancementLevel: 'medium'

Aggressive

Best for: Noisy environments, public spacesMaximum noise removal that prioritizes voice isolation over natural sound.
enhancementLevel: 'aggressive'

​
Noise Gate

Software-based noise gate filters out background noise between speech segments.

​
How It Works

The noise gate acts like an automatic mute button, opening only when you speak and closing during silence to eliminate ambient noise. 
const config = {
  audioConfig: {
    noiseGate: {
      enabled: true,
      threshold: 0.04,      // Volume threshold (0-1)
      attackTime: 30,       // How fast gate opens (ms)
      holdTime: 400,        // Hold open during pauses (ms)
      releaseTime: 150,     // How fast gate closes (ms)
    },
  },
};

​
Configuration Parameters

​
threshold
number
default:0.04
Volume level (0-1) required to open the gate. Lower values are more sensitive.
  • 0.02-0.03: Very sensitive, good for quiet environments
  • 0.04-0.06: Standard setting for most environments
  • 0.07-0.10: Less sensitive, for noisy environments
​
attackTime
number
default:30
Time in milliseconds for the gate to fully open when speech is detected.
  • 10-20ms: Very fast, may clip beginning of words
  • 30-50ms: Standard, natural sounding
  • 60-100ms: Slower, smoother transitions
​
holdTime
number
default:400
Time in milliseconds to keep gate open after speech stops, preventing choppy audio.
  • 200-300ms: Quick release, good for fast conversations
  • 400-500ms: Standard, handles normal pauses
  • 600-800ms: Longer hold, better for thoughtful speech
​
releaseTime
number
default:150
Time in milliseconds for the gate to fully close after hold time expires.
  • 50-100ms: Fast close, may sound abrupt
  • 150-200ms: Standard, natural fade
  • 250-400ms: Slow close, very smooth

​
Environment Presets

Pre-configured audio settings optimized for common environments. 
// Quiet office environment
{
  audioConfig: {
    environmentPreset: 'office',
  },
}

// Automatically configures:
// - Light noise suppression
// - Low noise gate threshold (0.03)
// - Standard timing parameters

​
Audio Debugging

Built-in tools to diagnose and fix audio issues.

​
Stutter Analyzer

Debug audio stuttering and performance issues: 
// Enable performance monitoring
const config = {
  debugAudioPerformance: true,
};

// In browser console:
audioStutterAnalyzer.setDebugMode(true);  // Start collection
audioStutterAnalyzer.getAnalysis();       // View analysis
audioStutterAnalyzer.clear();             // Clear buffer

// The analyzer tracks:
// - Audio underruns (stuttering)
// - Long DOM captures during audio
// - Buffer statistics
// - Correlation between renders and stutters

​
Common Issues and Solutions

  • Echo Problems
  • Choppy Audio
  • Background Noise
  • Muffled Voice
Issue: Hearing echo or feedbackSolutions:
// Enable echo cancellation
{
  audioConfig: {
    noiseSuppression: {
      enabled: true,
      // Echo cancellation is automatic
    },
  },
}
Additional Tips:
  • Use headphones when possible
  • Reduce speaker volume
  • Increase distance between mic and speakers

​
Best Practices

Test in Target Environment: Always test audio settings in the actual environment where the agent will be used
Start with Presets: Use environment presets as a starting point, then fine-tune if needed
Monitor Performance: Enable debug mode during development to catch audio issues early
User Feedback: Provide visual feedback for volume levels and mute status
Fallback Options: Have a text input fallback for environments where audio isn’t suitable

​
Advanced Configuration

​
Combining Multiple Techniques

Layer different audio processing techniques for optimal results: 
const config = {
  audioConfig: {
    // Start with a preset
    environmentPreset: 'office',
    
    // Override specific settings
    noiseSuppression: {
      enabled: true,
      enhancementLevel: 'medium',
    },
    
    // Fine-tune noise gate
    noiseGate: {
      enabled: true,
      threshold: 0.045,
      holdTime: 450,
    },
  },
};

​
Dynamic Adjustment

Adjust audio settings based on user feedback or environment detection: 
function adjustAudioForEnvironment(noiseLevel: 'low' | 'medium' | 'high') {
  const presets = {
    low: { environmentPreset: 'office' },
    medium: { environmentPreset: 'home' },
    high: { environmentPreset: 'noisy' },
  };
  
  // Update configuration dynamically
  updateAgentConfig({
    audioConfig: presets[noiseLevel],
  });
}

​
Platform Considerations

Audio processing behavior may vary across different platforms and browsers. Always test on your target platforms.

​
Browser Support

FeatureChromeFirefoxSafariEdge
Native Suppression✅✅✅✅
Echo Cancellation✅✅✅✅
Noise Gate✅✅✅✅
Koala AI✅✅⚠️✅
Safari may require additional permissions for advanced audio processing features.

VAD Configuration

Configure voice activity detection settings

Performance Optimization

Optimize audio processing for better performance
⌘I