hey! I’m getting a bug with the elevenlabs plugin where, when I have an LLM call that generates speech (fed to a session.say), then makes a tool call, then generates more speech (fed into the same session.say generator), ElevenLabs only reads out the first speech (and sometimes a little of the second). For Cartesia, this same flow works fine. I’m wondering if — 1. I’m not expected to pass multiple segments into session.say? 2. There’s a bug in the elevenlabs plugin?

I'm having a bit of trouble following the flow, is there a minimal example to reproduce? I don't see any known issues that are identical, only similar ones like this: https://github.com/livekit/agents/issues/3133 Also recommend you upgrade to the latest version of

livekit-agents

if you haven't already done so.

Gotcha. Yes, I made an issue here: https://github.com/livekit/agents/issues/4221 The problem is just — if I create a session.say with a generator, and that generator: 1. passes a chunk of text 2. pauses briefly 3. passes another chunk of text The second chunk of text never gets read. This is not the case with Cartesia.

Great, I see Long is addressing this within the issue