🎤 Micdrop

MicRecorder

The MicRecorder class provides functionality for recording audio from a microphone with voice activity detection (VAD). It uses the browser’s MediaRecorder API and integrates with a VAD (Voice Activity Detection) system for speech detection.

Overview

The MicRecorder class is a core component that can be used in two ways:

  1. As part of MicdropClient: The MicRecorder is automatically managed by MicdropClient when using it for voice conversations. MicdropClient creates an instance internally and handles all the microphone setup, speech detection, and audio streaming.

  2. As a standalone component: You can use MicRecorder directly if you only need microphone recording and speech detection functionality without the WebSocket communication and conversation management provided by MicdropClient.

This flexibility allows you to either use the full voice conversation capabilities through MicdropClient, or implement your own custom audio handling using just the microphone recording features of MicRecorder.

Features

  • Voice activity detection (VAD)
  • Multiple audio format support (ogg, webm, mp4, wav)
  • Event-based architecture
  • State management

Usage Example

import { Mic, MicRecorder } from '@micdrop/web'


// Create a new recorder instance with VAD config (string, VAD instance, or array)
const recorder = new MicRecorder('volume')


// Start the microphone, then listen to it
await Mic.start()
await recorder.start(Mic)


// Listen for events
recorder.on('StartSpeaking', () => {
  console.log('User started speaking')
})


recorder.on('StopSpeaking', () => {
  console.log('User stopped speaking')
})


recorder.on('Chunk', (chunk: Int16Array) => {
  // 100 ms of mono PCM16 at 16 kHz, ready to be sent to the server
  console.log('Received audio chunk:', chunk.length)
})

State

The recorder maintains a state object with the following properties:

interface MicRecorderState {
  isStarting: boolean // Whether the recorder is in the process of starting
  isStarted: boolean // Whether the recorder is currently active
  isSpeaking: boolean // Whether speech is currently detected
}

Events

The recorder emits the following events:

  • Chunk: Emitted when a new audio chunk is available, 100 ms of mono PCM16 at 16 kHz
  • StartSpeaking: Emitted when speech is detected
  • StopSpeaking: Emitted when speech ends
  • StateChange: Emitted when the recorder’s state changes

Methods

constructor(vadConfig?: VADConfig)

Creates a new MicRecorder instance with the provided VAD config. The config can be:

  • A string ('volume' or 'silero')
  • A VAD instance
  • An array of VAD configs for multiple VADs

start(mic: MicSource): Promise<void>

Starts the recorder, listening to the microphone it is given.

stop(): void

Stops the recorder and cleans up resources.

VAD

The MicRecorder class uses a VAD (Voice Activity Detection) system to detect speech. You can access the internal VAD instance via recorder.vad to update options or listen to VAD-specific events.

See VAD documentation for more information and available options.

Technical Details

  • Uses a delayed stream to avoid cutting off speech at the beginning of detection
  • Audio is recorded in chunks of 100ms when speech is detected
  • Default audio settings: 128kbps bitrate
  • Supports multiple audio formats with fallback options