Diarization (Speaker Identification)

import { createDiarization, assetModelPath } from 'react-native-sherpa-onnx/diarization';
import { createSTT } from 'react-native-sherpa-onnx/stt';

// Create diarization engine
const diarization = await createDiarization({
  modelPath: assetModelPath('models/speaker-embedding-model'),
});

// Analyze audio file
const result = await diarization.diarizeFile('/path/to/meeting.wav');

console.log('Number of speakers:', result.numSpeakers);
console.log('Speaker segments:', result.segments);
// [
//   { speakerId: 0, start: 0.0, end: 5.2 },
//   { speakerId: 1, start: 5.2, end: 8.7 },
//   { speakerId: 0, start: 8.7, end: 12.3 },
//   ...
// ]

// Combine with transcription
const stt = await createSTT({ /* ... */ });
const transcription = await stt.transcribeFile('/path/to/meeting.wav');

// Merge diarization with transcription
const transcript = mergeDiarizationWithTranscription(
  transcription,
  result.segments
);

console.log('Transcript with speakers:');
for (const segment of transcript) {
  console.log(`Speaker ${segment.speakerId}: ${segment.text}`);
}

await diarization.destroy();
await stt.destroy();

import { createDiarization } from 'react-native-sherpa-onnx/diarization';

const diarization = await createDiarization({ /* ... */ });

// Extract speaker embedding from enrollment audio
const aliceEmbedding = await diarization.extractEmbedding(
  '/path/to/alice-voice.wav'
);

const bobEmbedding = await diarization.extractEmbedding(
  '/path/to/bob-voice.wav'
);

// Register known speakers
await diarization.registerSpeaker('Alice', aliceEmbedding);
await diarization.registerSpeaker('Bob', bobEmbedding);

// Identify speakers in new audio
const result = await diarization.identifySpeakers(
  '/path/to/conversation.wav'
);

for (const segment of result.segments) {
  console.log(`${segment.speakerName}: [${segment.start}s - ${segment.end}s]`);
}
// Output:
// Alice: [0.0s - 5.2s]
// Bob: [5.2s - 8.7s]
// Alice: [8.7s - 12.3s]

await diarization.destroy();

import { createDiarization } from 'react-native-sherpa-onnx/diarization';
import { createStreamingSTT } from 'react-native-sherpa-onnx/stt';
import { createPcmLiveStream } from 'react-native-sherpa-onnx/audio';

const diarization = await createDiarization({ /* ... */ });
const stt = await createStreamingSTT({ /* ... */ });
const stream = await stt.createStream();

const mic = createPcmLiveStream({ sampleRate: 16000 });

let currentSpeaker = 0;

mic.onData(async (samples, sampleRate) => {
  // Detect speaker changes
  const speakerResult = await diarization.detectSpeaker(
    samples,
    sampleRate
  );
  
  if (speakerResult.speakerId !== currentSpeaker) {
    console.log(`Speaker changed: ${currentSpeaker} -> ${speakerResult.speakerId}`);
    currentSpeaker = speakerResult.speakerId;
  }
  
  // Transcribe with speaker label
  const { result } = await stream.processAudioChunk(samples, sampleRate);
  if (result.text) {
    console.log(`Speaker ${currentSpeaker}: ${result.text}`);
  }
});

await mic.start();

interface DiarizationResult {
  numSpeakers: number;
  segments: DiarizationSegment[];
}

interface DiarizationSegment {
  speakerId: number;
  speakerName?: string;
  start: number;  // seconds
  end: number;    // seconds
  confidence?: number;
}

interface SpeakerEmbedding {
  vector: number[];
  dimension: number;
}