The Vapi assistant I created has started saying "laughs" or "chuckles" sometimes, especially when you ask it to tell a joke. This is not wanted, and I can't even seem to write a prompt to stop it. How can I get this behavior to stop? I'm trying this in the prompt. VOICE CONTROL Do not include any words in your response about laughing. Specifically, when responding to a request for humor, never say laughs or chuckles or laugh softly. But I'm still getting this: User Uh, I would like you to tell us a joke, please. Can you do that? Assistant Chuckles warmly. Sure thing. I'd be happy to share a joke with you. Are you ready? Okay. Here goes. Why did the tomato turn red? Because it saw the salad dressing. Pauses for effect, then laughs good naturedly. Call ID: 553c1c7d-84c4-41ee-a6ed-ce39f18c67be I've tried switching LLMs (openAI and Anthropic both do this now). I'm using 11Labs.

@Nep It seems like the assistant is interpreting the instructions to include non-verbal expressions as part of the storytelling or joke-telling experience. To correct this, you might need to adjust the configuration of your voice assistant to exclude these non-verbal cues explicitly. Here's a more direct prompt that you could try: VOICE CONTROL When telling a joke or responding to humor-related requests, do not include any non-verbal expressions such as "laughs," "chuckles," or any other sound effects that imply laughter. Keep the response verbal and straightforward without any additional commentary or actions. If this does not resolve the issue, it may be necessary to look into the specific settings of the voice assistant or the platform you are using, such as 11Labs, and see if there are any additional configurations that control the inclusion of non-verbal expressions. If you continue to experience difficulties, please reach out to the support team of the voice API or the LLM provider for further assistance.

Thanks. I think that did it.

Awesome.