This message was deleted.

It seems likely that your issue might be related to the audio format or the way the audio is being streamed to LiveKit. According to the LiveKit documentation, plugins typically handle audio in specific formats, and if your server is sending audio in MP3, it might not be compatible with the expected format. LiveKit's TTS plugins often use

SynthesizedAudio

which expects audio frames in a format like PCM, not MP3. To build a custom TTS plugin for LiveKit, you can subclass the

agents.tts.TTS

class. This involves implementing the

synthesize

method to convert text to audio frames that LiveKit can handle. You should ensure that the audio is in a compatible format, such as PCM, and use the

ChunkedStream

or

SynthesizeStream

to manage the audio data. You can refer to existing plugins like ElevenLabs for structure and implementation details. Would you like more detailed guidance on implementing this? Sources: - ElevenLabs 👍 0 button 👎 0 button 💡 Hint: Mention @runllm in the thread for followups.