src/app/api/generators/speech/generateVoiceWithElevenLabs.txt

import { getMediaInfo } from "../../utils/getMediaInfo"
import { readMp3FileToBase64 } from "../../utils/readMp3FileToBase64"

export async function generateSpeechWithElevenLabs({
  text,
  audioId,
  debug = false,
}: {
  text: string
  audioId: string
  debug?: boolean
}): Promise<{
  filePath: string
  fileName: string
  format: string // "mp3"
  base64: string // data uri
  durationInSec: number
  durationInMs: number
}> {
  const api = await ElevenLabs()

  // Converts text to speech, saves the file to the output folder and returns the relative path to the file.
  // Output file is in the following format: TTS_date-time.mp3
  // Returns an object with the following structure: { code: CODE, message: "STATUS_MESSAGE" }
  const result = await api.tts(
    text,
    audioId
  )

  // ...really? that's the API?
  let relativeOutputPath = result.message.split("File written successfully:").pop().trim()

  // we remove the ./ at the beginning, so we get something like:
  // "/../../../../var/folders/x4/2w7-------------------"
  // then we remove relative navifation to only keep this:
  // "/var/folders/x4/2w7-------------------"
  const filePath = relativeOutputPath.slice(1).replaceAll("/..", "")

  const fileName = filePath.split("/").pop()

  const format = fileName.split(".").pop()

  const { durationInSec, durationInMs } = await getMediaInfo(filePath)
  
  const base64 = await readMp3FileToBase64(filePath)
  
  return {
    filePath,
    fileName,
    format,
    base64,
    durationInSec,
    durationInMs,
  }
}