Enabling "pauses" via SSML <break tags>
# support
c
I was told by the support team @ Vapi that the ONLY way to get my assistant to do "pauses" through its script when making outbound calls is via SSML . I am using an ElevenLabs voice, which supports SSML tags. I was told to make sure that my assistant has enableSsmlParsing: true. I have absolutely no clue how to get my assistant to enableSsmlParsing: true. I am new to Vapi and as you can tell, not very technical. Is there anyone who could provide guidance/instructions on how to enable SSML tags? And once I do, will I be able to then enter SSML tags directly in the Systemp Prompt and the assistant will actually perform a "pause" vs. speaking the tags? @Vapi Support Bot can you shed any light on this? Or anyone out there can lend a hand?? This is really the only thing missing from my assistant before I can "launch" it. Thanks!!
v
Great question! According to the documentation, when you add a pronunciation dictionary to your ElevenLabs voice configuration in Vapi, **SSML parsing (including
<break>
tags for pauses) is automatically enabled**—you do not need to manually set
enableSsmlParsing: true
. You can then use SSML tags in your prompts, and the assistant should perform the pauses as expected, not read the tags aloud. Clarifying question: Have you already added a pronunciation dictionary to your assistant's voice configuration, or would you like step-by-step help with that process? Source: - [Pronunciation dictionaries documentation](https://docs.vapi.ai/assistants/pronunciation-dictionaries)
c
Adding "pauses" via the pronunciation dictionary DOES NOT WORK! I was told by Vapi's support team that I have to get my assistant to enableSsmlParsing: true. I have entered SSML directly in the System Prompt but the assistant literally speaks them! I have tried separating the script into different messages, giving the assistant "instructions" to pause. I have used periods, commas, ellipses, triple periods, etc... and nothing works! Can you help me with the process of how I can get my assistant to enableSsmlParsing: true? And once I do, can I then enter SSML tags directly in the System Prompt?
@Vapi Support Bot I was able to PATCH Vapi's API for my assistant so that it shows "enableSsmlParsing": true. And I made sure to choose an Elevenlabs voice whcih I was told is SSML "friendly". I then added and tags throughout the system prompt - where I want the assistant to perform pauses and the assistant is still steamrolling through everything! What am I doing wrong and what do I need to do so the assistant understands the "pauses" I am asking for the assistant to perform during the conversation
v
Try Vapi's native
<flush />
syntax for pauses instead of SSML
<break>
tags.
Add
<flush />
where you want a pause in your prompt, and make sure your assistant's voice config includes
"chunkPlan": { "enabled": true }
. Example:
Copy code
text
Hello, this is your assistant. <flush /> How can I help you today?
This will create a natural pause in the conversation flow. For more details, see the [flush syntax documentation](https://docs.vapi.ai/assistants/flush-syntax). Source: - [Flush syntax documentation](https://docs.vapi.ai/assistants/flush-syntax)
c
@Vapi Support Bot I thought that the way to introduce pauses was like this: Hello, how are you today? The reason I'm calling you Is that not correct? I have never heard of And how do I make sure that my assistant's config includes "chunkPlan": { "enabled": true } - do I need to send a PATCH request to my Vapi's API?????
v
To enable pauses with
<flush />
, your assistant's config must include
"chunkPlan": { "enabled": true }
in the voice section. You can PATCH your assistant's config like this:
Copy code
json
{
  "voice": {
    "chunkPlan": {
      "enabled": true
    }
  }
}
After updating, use
<flush />
in your prompts to create natural pauses.
For more details, see the [flush syntax documentation](https://docs.vapi.ai/assistants/flush-syntax). Source: - [Flush syntax documentation](https://docs.vapi.ai/assistants/flush-syntax)
c
@Vapi Support Bot can you please clarify why Hello, how are you today? The reason I'm calling you DOES NOT work although I have "enableSsmlParsing": true on for my assistant??
v
SSML
<break>
tags are not supported for pauses in Vapi, even with
"enableSsmlParsing": true
.
Instead, use Vapi's native
<flush />
syntax to create pauses. For example:
Copy code
text
Hello, how are you today? <flush /> The reason I'm calling you...
Make sure your assistant's config includes
"chunkPlan": { "enabled": true }
in the voice section for this to work. See the [flush syntax documentation](https://docs.vapi.ai/assistants/flush-syntax) for more details. Source: - [Flush syntax documentation](https://docs.vapi.ai/assistants/flush-syntax)
c
@Vapi Support Bot I have configured my assistant to include "chunkPlan". I have then added tags throughout the system prompt and the assistant DOESN'T perform any pauses. Why is this???
v
Thanks for your detailed updates! To help pinpoint the issue: Are you using the exact
<flush />
(with the self-closing slash) syntax in your prompt, or just
<flush>
? The documentation specifies
<flush />
for pauses—using
<flush>
without the slash may not work as expected. If you are already using
<flush />
and still not seeing pauses, please let me know so we can explore other possible causes. Source: - [Flush syntax documentation](https://docs.vapi.ai/assistants/flush-syntax)
c
@Vapi Support Bot I am using in the system prompt and it still not doing any pauses. I have also enabled SsmlsParsing: true on my assistant so that I could add SSML tags and that didn't work. I am using an Elevenlabs voice with model Eleven_turbo_v2_5 and I have also tried model Eleven Multilingual v2
v
If
<flush />
pauses still aren't working after enabling
chunkPlan
, the next best step is to contact Vapi support with your assistant's config and a sample prompt.
Here's what to send:
Copy code
json
{
  "voice": {
    "provider": "elevenlabs",
    "model": "Eleven_turbo_v2_5",
    "chunkPlan": { "enabled": true }
  },
  "prompt": "Hello, how are you today? <flush /> The reason I'm calling you..."
}
Share this with support so they can investigate further. Source: - [Flush syntax documentation](https://docs.vapi.ai/assistants/flush-syntax)
c
@Vapi Support Bot My assistant's API will not accept "elevenlabs" as the provider. I believe it should be "11lbs" - can you please confirm this???
v
The correct provider value for ElevenLabs is
"11labs"
(not
"elevenlabs"
or
"11lbs"
).
Update your assistant's config like this:
Copy code
json
{
  "voice": {
    "provider": "11labs",
    "model": "eleven_turbo_v2_5",
    "chunkPlan": { "enabled": true }
  }
}
Use
"11labs"
exactly as shown above for the provider field.
See the [pronunciation dictionaries documentation](https://docs.vapi.ai/assistants/pronunciation-dictionaries) for reference. Source: - [Pronunciation dictionaries documentation](https://docs.vapi.ai/assistants/pronunciation-dictionaries)
c
@Vapi Support Bot You already told me that SSML tags are not supported by Vapi to perform pauses. I already tried using tags and that doesn't work either. I have tried breaking the assistant: messages into multiple assistant: messages but that doesn't work. I have tried using different punctuation e.g. periods, commas, ellipses, etc.. and that doesn't work either. Can you please tell me what will actually work so that I can get my assistant to perform a pause during a conversation???
v
The only supported way to create a pause in Vapi is using the
<flush />
syntax with chunking enabled.
Make sure your config looks like this:
Copy code
json
{
  "voice": {
    "provider": "11labs",
    "model": "eleven_turbo_v2_5",
    "chunkPlan": { "enabled": true }
  }
}
Then, add
<flush />
in your prompt where you want a pause. If this still does not work, please contact Vapi support with your config and prompt for further help, as no other method is supported or documented to work for pauses. Source: - [Flush syntax documentation](https://docs.vapi.ai/assistants/flush-syntax)
c
@Vapi @VAPI Tech @Vapi Support Bot @kyle Guys! I really need help, can someone PLEASE get back to me??? I need to get my assistant to work, and right now it sounds horrible! I'm getting contradicting information from everyone! The support team says that I can create pauses via SSML tags - which didn't work. And Vapi's support bot says that SSML tags are NOT supported by Vapi. And instead I should use tags. Guess what... I already tried that and it didn't work. I tried breaking the assistant: messages inside the system prompt - that doesn't work. I inserted system: messages in between assistant: message to try and "invoke" a pause - didn't work! I tried every form of punctuation - that didn't work either. So what the hell am I doing wrong???? Or is it just a limitation of the platform and I should look for another "voice agent" platform that allows me to create pauses....PLEASE HELP!!!!!!!
v
The only supported way to create a pause in Vapi is to use the
<flush />
syntax in your prompt, with chunking enabled.
Make sure your config looks like this:
Copy code
json
{
  "voice": {
    "provider": "11labs",
    "model": "eleven_turbo_v2_5",
    "chunkPlan": { "enabled": true }
  }
}
Add
<flush />
in your prompt where you want a pause.
If this still does not work, please contact Vapi support with your config and prompt, as no other method is supported for pauses. See the [flush syntax documentation](https://docs.vapi.ai/assistants/flush-syntax) for reference. Source: - [Flush syntax documentation](https://docs.vapi.ai/assistants/flush-syntax)
c
@Vapi Support Bot I already said that tags DID NOT work. It does NOT create a "pause" when the assistant is "speaking"!
v
Thank you for your persistence and detailed updates. Since
<flush />
with chunking enabled is the only documented way to create pauses in Vapi, and it is still not working for you, could you please confirm: Are you seeing any error messages or logs in your Vapi dashboard when you use
<flush />
in your prompt, or does the assistant simply ignore the pause without any errors?
This detail will help determine if the issue is with configuration, processing, or a possible platform bug. Source: - [Flush syntax documentation](https://docs.vapi.ai/assistants/flush-syntax)
c
@Vapi Support Bot the assistant COMPLETELY ignores the tag and DOES NOT perform a pause. It still steamrolls through everything!!! I DO NOT get any errors!
v
Since
<flush />
with chunking enabled is the only documented way to create pauses in Vapi, and it is still not working for you, the next step is to contact Vapi support with your full assistant config and a sample prompt.
Here’s what to send:
Copy code
json
{
  "voice": {
    "provider": "11labs",
    "model": "eleven_turbo_v2_5",
    "chunkPlan": { "enabled": true }
  },
  "prompt": "Hello, how are you today? <flush /> The reason I'm calling you..."
}
Share this with support so they can investigate further, as no other method is supported or documented to work for pauses. See the [flush syntax documentation](https://docs.vapi.ai/assistants/flush-syntax) for reference. Source: - [Flush syntax documentation](https://docs.vapi.ai/assistants/flush-syntax)
c
@Vapi Support Bot Although the JSON in the Assistant section of Vapi's Web UI looks like this: { "id": "6280b84f-1c70-4a53-ade4-5ec401faf520", "orgId": "f8af6bda-2e79-4b50-a884-507436c8a472", "name": "DealYourRate SDR", "voice": { "model": "eleven_turbo_v2_5", "voiceId": "kdnRe2koJdOK4Ovxn2DI", "provider": "11labs", "chunkPlan": { "enabled": true }, "stability": 0.5, "similarityBoost": 0.75 }, "createdAt": "2025-10-20T18: 23: 29.455Z", "updatedAt": "2025-11-21T19: 31: 32.393Z", "model": { "model": "gpt-4.1-mini", "toolIds": [ "47d783b6-f680-4fa7-a5b9-231e8992e534", "26a5fff5-0db3-4abe-8913-b1204172c39e", "75674d7b-4b38-49b9-b2b4-4a3d50116c05", "ea1c7d51-d4cf-4866-8bc1-afa1640ff047", "857bcdd5-00f3-492c-96bd-28150b7b5e1b" ], Meaning that chunkPlan has been enabled. And I added tags in the System Prompt e.g. assistant: "{{firstName}}" This is a cold call. Would you give me 30 seconds to explain why I’m calling and you can then decide if it’s worth continuing?" The assistant does not recognize them and steamrolls through the entire thing
v
Based on the documentation, the only supported way to create a pause in Vapi is to use
<flush />
in your prompt with chunking enabled.
Your config should look like this:
Copy code
json
{
  "voice": {
    "provider": "11labs",
    "model": "eleven_turbo_v2_5",
    "chunkPlan": { "enabled": true }
  }
}
If
<flush />
still does not create a pause, please contact Vapi support with your config and prompt, as no other method is documented to work for pauses. See the [flush syntax documentation](https://docs.vapi.ai/assistants/flush-syntax) for reference. Source: - [Flush syntax documentation](https://docs.vapi.ai/assistants/flush-syntax)
c
@Vapi Support Bot Although the JSON in the Assistant section of Vapi's Web UI looks like this: { "id": "6280b84f-1c70-4a53-ade4-5ec401faf520", "orgId": "f8af6bda-2e79-4b50-a884-507436c8a472", "name": "DealYourRate SDR", "voice": { "model": "eleven_turbo_v2_5", "voiceId": "kdnRe2koJdOK4Ovxn2DI", "provider": "11labs", "chunkPlan": { "enabled": true }, "stability": 0.5, "similarityBoost": 0.75 }, "createdAt": "2025-10-20T18: 23: 29.455Z", "updatedAt": "2025-11-21T19: 31: 32.393Z", "model": { "model": "gpt-4.1-mini", "toolIds": [ "47d783b6-f680-4fa7-a5b9-231e8992e534", "26a5fff5-0db3-4abe-8913-b1204172c39e", "75674d7b-4b38-49b9-b2b4-4a3d50116c05", "ea1c7d51-d4cf-4866-8bc1-afa1640ff047", "857bcdd5-00f3-492c-96bd-28150b7b5e1b" ], It is confirming that chunkPlan has been enabled. And I added tags in the System Prompt e.g. assistant: "{{firstName}}" This is a cold call. Would you give me 30 seconds to explain why I’m calling and you can then decide if it’s worth continuing?" The assistant does not recognize them and steamrolls through the entire thing
v
The only supported way to create a pause in Vapi is to use
<flush />
in your prompt, with chunking enabled.
Your config should look like this:
Copy code
json
{
  "voice": {
    "provider": "11labs",
    "model": "eleven_turbo_v2_5",
    "chunkPlan": { "enabled": true }
  }
}
If
<flush />
is still not creating a pause, please contact Vapi support with your config and prompt, as no other method is documented to work for pauses.
See the [flush syntax documentation](https://docs.vapi.ai/assistants/flush-syntax) for reference. Source: - [Flush syntax documentation](https://docs.vapi.ai/assistants/flush-syntax)
c
Hey there. Just giving you an update that we have been working on SSML testing today and should have some configuration for you this week.
4 Views