Not getting desired result with speech-to-text API

TensorFlow
1

Howdy. Im trying to use speech-to-text API on an audio file that is 02:22 long and 1,4 MB.

Im using options like so:

const filename = "audio.mp3";
  const encoding = "mp3";
  const sampleRateHertz = 16000;
  const languageCode = "en-US";
  const model = "phone_call";

  const config = {
    encoding: encoding,
    sampleRateHertz: sampleRateHertz,
    languageCode: languageCode,
    model: model,
    enableAutomaticPunctuation: true,
  };

But my result is not that great tbh. Many words are very bad match of voice.
Anyone know any possible way to improve the speech-to-text?

Im trying to transcribe a audio file from a conversation between 2 people.
The general audio quality is great and clear.