Hi CloudHeads,
Your guidance is mostly accurate. When using function calls in Retell AI, the agent automatically waits for the function to return results before continuing the conversation—no manual pauses are required for that workflow(1).
For adding pauses in agent speech (when not waiting on a function), Retell AI officially supports using spaced dashes(1). You can control speech pacing by adding - with spaces around them in the agent’s response(1):
The number is 2 - 1 - 3 - 4
For longer pauses, you can add multiple dashes with double spaces between them(1):
The number is 2 - - - - 1 - - - - 3
Regarding the <break time="1.0s" /> SSML tag you mentioned, this approach has been discussed in community channels(2)(3). Some users report success with ElevenLabs SSML tags like <break time="3s" />(2), though the official documentation focuses on the spaced dash method(1). The maximum pause supported by ElevenLabs SSML is 3.5 seconds(3).
If you’re experiencing issues with pauses not working as expected, note that some users have reported the dash method causing unexpected sounds with certain TTS providers and voices(4). In those cases, wrapping the break tag in a <speak> tag or adjusting the prompt to use natural language pauses has helped(4).
For the specific use case of checking “Jan’s availability,” if you’re using a function call, the agent will naturally wait. If you want to add a conversational pause while the function executes, you could prompt the agent to say something like “Let me check that for you” before the function call, then use the dash method or SSML tags for any additional pauses needed in the response.