transcribe.ts

import fs from 'fs';
import { createClient } from '@deepgram/sdk';

interface Sentence {
  text: string;
  start: number;
  end: number;
}

interface Word {
  word: string;
  start: number;
  end: number;
  punctuated_word?: string;
}

export interface TranscriptionResult {
  sentences: Sentence[];
  words: Word[];
}

function getDeepgramClient() {
  return createClient(process.env.DEEPGRAM_API_KEY!);
}

export async function transcribeDummy(
  audioPath: string
): Promise<TranscriptionResult> {
  console.log(
    `[transcribeDummy] Using static transcription file instead of transcribing ${audioPath}`
  );

  const transcriptFile = await fs.promises.readFile('transcript.json', 'utf8');
  const transcription = JSON.parse(transcriptFile) as TranscriptionResult;

  return transcription;
}

export async function transcribe(
  audioPath: string
): Promise<TranscriptionResult> {
  const deepgramClient = getDeepgramClient();
  const { result, error } =
    await deepgramClient.listen.prerecorded.transcribeFile(
      fs.createReadStream(audioPath),
      {
        model: 'nova-3',
        smart_format: true,
        punctuate: true,
        dictation: true,
        utterances: true,
        diarize: true,
        language: 'fr',
        filler_words: true,
      }
    );

  if (error) {
    throw error;
  }

  // get the sentences and words from the result
  const sentences: Sentence[] = [];
  for (const paragraph of result.results.channels[0].alternatives[0].paragraphs
    ?.paragraphs ?? []) {
    for (const sentence of paragraph.sentences) {
      sentences.push({
        text: sentence.text,
        start: sentence.start,
        end: sentence.end,
      });
    }
  }
  const words: Word[] = [];
  for (const word of result.results.channels[0].alternatives[0].words) {
    words.push({
      word: word.word,
      start: word.start,
      end: word.end,
      punctuated_word: word.punctuated_word,
    });
  }
  return {
    sentences,
    words,
  };
}