Hi, I need help with two issues: 1) The assistant is "blazing" through the pitch in the System Prompt. It totally sounds scripted and sounds like a robot. How can I introduce a "pause" so that it sounds and feels more "human-like"? I already did the ellipsis (...) and nothing. I also split the pitch into different "prompts" within the System Prompt and nothing i.e. system: The prospect has just agreed to give you 30 seconds. assistant: "Thanks {{firstName}}, I’ll be quick." system: You have thanked them. Now, deliver the first part of the pitch. assistant: "Deal Your Rate is......." system: The AI has just delivered Part 1 of the pitch. The AI must now go silent and wait for the user's next utterance (e.g., "Okay," "Uh-huh," "Go on"). 2) I want the AI to hang up. It doesn't, it just sticks around... A) I added assistant: [function:endCall()] on the prompt, but that doesn't work B) Added "Have a great day!" @ End Of Message - still doesn't hang up!

Hey, I’ve worked on this before — the “robotic” tone happens because VAPI ignores ellipses, so you’ll need SSML tags to create natural pauses. The hang-up issue usually means endCall() isn’t being triggered by the right event. I can help you configure both pacing and proper call termination. Are you using the default VAPI TTS or a custom voice engine? @Charlie

Hi @Vercel I'm new to this so please "hold hands" with me, hehe! I am using a voice from 11Labs and the transcriber is Deepgram - does that help? Please let me know what else you need to do and how to apply SSML tags into the System Prompt in order to create natural pauses. And how to ensure it does hang. Happy to share the entire System Prompt. I can also do a video call - whatever is easiest for you. I look forward to hearing back from you! Thanks!

Hi Charlie, To introduce pauses and make the assistant sound more human-like, consider using provider-specific pause characters. For example, with Twilio, use

w

for a 0.5-second pause or

W

for a 1-second pause. This can help slow down the pace. Additionally, ensure that the system prompt is structured to encourage natural pauses. If needed, you can also explore the

startSpeakingPlan

configuration to manage speech cadence, especially at the beginning of calls. For additional guidance, you might want to refer to the [IVR Navigation](https://docs.vapi.ai/ivr-navigation) document.

Thanks for your response! I have already structured the System Prompt to encourage natural pauses. Most of the pauses I'm looking for the assistant to do are more during the middle and towards the end of the "script". Based on the URL you sent over, the startSpeakingPlan is meant for the beginning of the call. And to be honest, I'm not sure how to implement "pauses" via Twilio - any guidance you can provide? Thanks!

To add a natural pause before the assistant speaks, update your voice pipeline configuration to include the

waitSeconds

parameter in the

startSpeakingPlan

. For example:

json
"startSpeakingPlan": {
  "waitSeconds": 0.4
}

This will insert a short pause before each assistant response, making the conversation sound more human-like. You can adjust the value (e.g., 0.4–0.8) for longer pauses as needed. For more details, see the [Vapi voice pipeline configuration guide](https://docs.vapi.ai/customization/voice-pipeline-configuration). Source: - [Voice pipeline configuration](https://docs.vapi.ai/customization/voice-pipeline-configuration)

@Vapi Support Bot what do you mean by "update your voice pipeline configuration to include waitSeconds parameter in the startspeaking plan?? Where do I insert it, in the System Prompt?

You should not add

waitSeconds

to the System Prompt. Instead, add it to your assistant's voice pipeline configuration under

startSpeakingPlan

. For example:

json
"startSpeakingPlan": {
  "waitSeconds": 0.4
}

This configuration is set in your assistant's settings (not in the prompt text). For more details, see the [Vapi voice pipeline configuration guide](https://docs.vapi.ai/customization/voice-pipeline-configuration). Source: - [Voice pipeline configuration](https://docs.vapi.ai/customization/voice-pipeline-configuration)

@Vapi Support Bot The pauses I want the assistant to do are not at the beginning or at the end of the "pitch". They are throughout the entire conversation.

Thanks for clarifying your goal. Based on the documentation, the

waitSeconds

parameter only adds a pause before each assistant response, not within the response itself. There is no documented way to insert mid-response pauses (like SSML

<break>

tags) when using 11Labs TTS via Vapi. Could you confirm if you specifically need mid-sentence pauses within a single assistant message, or would a pause before each response (using

waitSeconds

) be sufficient for your use case? This will help determine if a workaround or different configuration is needed. Source: - [Voice pipeline configuration](https://docs.vapi.ai/customization/voice-pipeline-configuration)

@Vapi Support Bot The pauses I want to generate are throughout the entire conversation the assistant would have with a prospect. For example, I want the assistant to say the first sentence of her pitch i.e. Acme corporation is a great construction company. I then want it to pause for 1 second and continue with the second sentence of the pitch i.e. We have been in business over 10 years. Otherwise the assistant runs both sentences together and sounds like it is steamrolling through the entire pitch. So based on this, am I able to insert pauses or not??

Based on the documentation, there is currently no way to insert mid-sentence pauses (like SSML

<break>

tags) within a single assistant message when using 11Labs TTS via Vapi—only a pause before each response is supported using the

waitSeconds

parameter in your voice pipeline configuration ([documentation](https://docs.vapi.ai/customization/voice-pipeline-configuration)). To help find the best possible workaround: Would splitting your pitch into multiple assistant responses (so each sentence is a separate response, with a pause before each) be an acceptable solution for your use case, or do you specifically require true mid-sentence pauses within a single message? Source: - [Voice pipeline configuration](https://docs.vapi.ai/customization/voice-pipeline-configuration)

@Vapi Support Bot I have already tried splitting the messages so that each sentence is in a separate response but that doesn't work either. For example: assistant: "Acme corporation is a great construction company" system: "You have said your first part of step 6. You MUST pause and move to the next part of your pitch assistant: "We have been around in business for over 10 years" Doing that doesn't work. It still stays both sentences one right after the other without a pause. So based on that, what options do I have with Vapi??

Thanks for your detailed explanation. Since the current Vapi + 11Labs setup does not support mid-sentence pauses, would you be open to exploring a different TTS provider (such as Azure or Google) that supports SSML

<break>

tags for more natural pacing, or do you need to stay with 11Labs for your project? Source: - [Voice pipeline configuration](https://docs.vapi.ai/customization/voice-pipeline-configuration)

@User I don't need to use 11Labs, I'm open to using another TTS provider. But how do I this within Vapi? And how would I then enable SSML tags?

To enable SSML

<break>

tags, add

"enableSsmlParsing": true

to your assistant's voice pipeline configuration under the TTS provider settings—not in the System Prompt or via n8n. For example:

json
"tts": {
  "provider": "azure",
  "enableSsmlParsing": true
}

Update this in your assistant's settings in the Vapi dashboard. For more details, see the [voice pipeline configuration guide](https://docs.vapi.ai/customization/voice-pipeline-configuration). Source: - [Voice pipeline configuration](https://docs.vapi.ai/customization/voice-pipeline-configuration)

@Vapi Support Bot You are saying that I can update this in your assistant settings. I am in the assistant settings and under "Voice Configuration", I do not see an option or a place where I can enter the JSON you provided to enable SSML. Can you please confirm where I need to enter it? Thanks!

Thanks for your follow-up. Based on the documentation,

"enableSsmlParsing": true

should be added to your assistant's voice pipeline configuration under the TTS provider settings in the Vapi dashboard—not in the System Prompt or via n8n. Since you do not see a place to enter this JSON in the "Voice Configuration" section, could you please confirm: Are you able to access an "Advanced" or "Custom JSON" editor within your assistant's settings, or is your dashboard limited to preset dropdowns and fields? This will help determine if your account has access to the necessary configuration options or if further support is needed. For reference, see the [Vapi voice pipeline configuration guide](https://docs.vapi.ai/customization/voice-pipeline-configuration). Source: - [Voice pipeline configuration](https://docs.vapi.ai/customization/voice-pipeline-configuration)

@Vapi Support Bot My dashboard is limited to preset dropdowns and fields. How am I able to access an "Advanced" or "Custom JSON" editor within my assistant's settings??

Currently, advanced speech configurations (like enabling SSML parsing) can only be set via API, and the dashboard does not offer an "Advanced" or "Custom JSON" editor if you only see preset dropdowns and fields. To proceed, you will need to contact Vapi support to request access to advanced configuration options or API setup for your assistant. For more details, see the [Vapi speech configuration documentation](https://docs.vapi.ai/assistants/speech-configuration). Source: - [Speech configuration](https://docs.vapi.ai/assistants/speech-configuration)

Same thing! Great apology but no answer to my problem. Can you please help me understand how to get my assistant to accepts SSML ? I have already been told that I need to get my assistant to enableSsmlParsing: true. But I have no clue how to enable this? And once I do, will I be able to enter SSML tags directly in the System Prompt???

DMed you