How to prevent splitting longer speech into 250 ms chunks in Next.js Livestream Starter?

[Hunaid98]

Hi there,

I am working on a real-time voice transcription project using the Next.js Livestream Starter, and I am facing an issue with how the microphone input is being handled. Specifically, the microphone only allows audio chunks of 250 ms, which causes longer speech to be split into small parts.

I would like to:

Prevent speech from being split into such small chunks (250 ms) and instead allow continuous recording or larger segments.
Accumulate the audio data and send it as one complete request once the user finishes speaking, rather than breaking it into multiple requests for each small chunk.
Current Approach:
I’m using MediaRecorder to capture the audio in my Next.js app, and I use a listener for the dataavailable event, but I am forced to set a timeslice argument (e.g., microphone.start(500)) which results in frequent breaks in the recording.

Here's the relevant code:

const setupMicrophone = async () => {
  setMicrophoneState(MicrophoneState.SettingUp);

  try {
    const userMedia = await navigator.mediaDevices.getUserMedia({
      audio: {
        noiseSuppression: true,
        echoCancellation: true,
      },
    });

    const microphone = new MediaRecorder(userMedia);

    setMicrophoneState(MicrophoneState.Ready);
    setMicrophone(microphone);
  } catch (err: any) {
    console.error(err);

    throw err;
  }
};

const startMicrophone = useCallback(() => {
  setMicrophoneState(MicrophoneState.Opening);

  if (microphone?.state === "paused") {
    microphone.resume();
  } else {
    microphone?.start(250);
  }

  setMicrophoneState(MicrophoneState.Open);
}, [microphone]);

Question:
How can I prevent the microphone from breaking the speech into 250 ms chunks?
Is there a way to handle continuous speech in a more efficient manner using the MediaRecorder API or any other approach within the Next.js Livestream?

Thanks in advance for any advice or recommendations!

[team-deepgram]

Thanks for asking your question about Deepgram! If you didn't already include it in your post, please be sure to add as much detail as possible so we can assist you efficiently, such as:

• The request_id if you have a question about your requests or transcription responses.
• The features you used or the full api.deepgram.com URL you sent your request to, including parameters.
• Any code snippets you can share.

[jkroll-deepgram]

Hi @Hunaid98, when using our real-time streaming endpoint, it's normal to send small chunks of audio to Deepgram. We concatenate them and process longer segments of audio to restore the context.

If you want to record a full audio file and then send it to Deepgram, that would use our pre-recorded transcription, not streaming.