No audio output on macOS with Elevenlabs TTS

[andypotato]

Using the Elevenlabs TTS Plugin I am unable to get any audio output on any client. It does work without problems on Windows. This most likely not a client issue but a server side issue specific to macOS.

• Running the voice agent on a windows system all connected client web browsers (mac, windows, ios, android) worked without any issue
• However running the voice agent on macOS, none of the clients were able to play any audio

There are no errors logged on server or client side.

Steps to reproduce

Creating a basic voice agent using the following code:

const voiceAgent = new pipeline.VoicePipelineAgent(
    vad, // created in prewarm proc.userData.vad = await silero.VAD.load();
    new deepgram.STT({ language: 'zh' }),
    new openai.LLM(),
    new elevenlabs.TTS({ modelID: 'eleven_turbo_v2_5' })
);

voiceAgent.start(room, participant);

Frontend is a minimal HTML / Javascript client. Relevant code to subscribe to the agent voice track (auto subscription is enabled):

handleTrackSubscribed(track, publication, participant) {
  if (track.kind === "audio") {
    const audioElement = track.attach();
    document.body.appendChild(audioElement);
    console.log(`Subscribed to audio track of ${participant.identity}`);
  }
}

[nbsp]

are you using the latest version of agents and rtc-node? we recently fixed an issue similar to this one

[andypotato]

After updating to the latest version of rtc-node the issue on mac is fixed