Function synthesizeWithDeepgram

apps/sim/app/api/tools/tts/unified/route.ts:367–431

(
  params: DeepgramTtsParams
)

}

async function synthesizeWithDeepgram(
  params: DeepgramTtsParams
): Promise<{ audioBuffer: Buffer; format: string; mimeType: string; duration?: number }> {
  const {
    text,
    apiKey,
    model = 'aura-asteria-en',
    encoding = 'mp3',
    sampleRate,
    bitRate,
    container,
  } = params

  const queryParams = new URLSearchParams({
    model: model,
    encoding: encoding,
  })

  if (sampleRate && encoding === 'linear16') {
    queryParams.append('sample_rate', sampleRate.toString())
  }

  if (bitRate) {
    queryParams.append('bit_rate', bitRate.toString())
  }

  if (container && container !== 'none') {
    queryParams.append('container', container)
  }

  const response = await fetch(`https://api.deepgram.com/v1/speak?${queryParams.toString()}`, {
    method: 'POST',
    headers: {
      Authorization: `Token ${apiKey}`,
      'Content-Type': 'application/json',
    },
    body: JSON.stringify({ text }),
  })

  if (!response.ok) {
    const error = await readTtsErrorJson(response, 'Deepgram TTS error response')
    const errorMessage = getTtsErrorMessage(error, response.statusText)
    throw new Error(`Deepgram TTS API error: ${errorMessage}`)
  }

  const audioBuffer = await readResponseToBufferWithLimit(response, {
    maxBytes: MAX_TTS_AUDIO_BYTES,
    label: 'Deepgram TTS audio response',
  })

  let finalFormat: string = encoding
  if (container === 'wav') {
    finalFormat = 'wav'
  } else if (container === 'ogg') {
    finalFormat = 'ogg'
  }

  const mimeType = getMimeType(finalFormat)