I've got an assistant with a

First Message

of

Hi. Thanks for calling...

. When I make a web call there are no issues, however when I call the number from a phone (Twilio number in Vapi) the

Hi

in the

First Message

is usually partially cut off. Is there any way to add a delay after the call is answered before the assistant speaks? Does anyone else have this issue? Thanks

@MS I understand the issue you're experiencing with the initial greeting being cut off when answering phone calls. This is a common occurrence with voice applications, especially when there's a slight delay in the connection establishment. To address this, you can add a brief pause before the assistant starts speaking. This can be done by inserting a short period of silence at the beginning of the

First Message

. Depending on the capabilities of your voice API, this could be achieved by adding a pause command or symbol, such as a comma or a special tag if your system supports SSML (Speech Synthesis Markup Language). For example, if your system supports SSML, you might add a delay like this:

xml
<break time="1s"/>

Or, if your system uses plain text with commas for brief pauses:

, Hi. Thanks for calling...

If you're unsure how to implement this or if the system supports such a feature, please ask support for more info on this. They can provide you with the specific method to add a delay based on the capabilities of your voice API.

Thanks. I tried both of these in the first message box with no real success. Would this need to be done in Vapi or in Twilio?

@MS Could you share both Call IDs using Eleven Flash V2.5 and Eleven Multilingual V2 Voice Models?

Thanks @User . I've included a few tests below, all done using the same assistant. While the Eleven Multilingual V2 is better generally than Flash, it is not error free. In the tests below, the word "Hi" is the first word in the

First Message

field. Let me know if you need any more info. Overall I think

Eleven Turbo V2

is the most consistently error free with the first word. Eleven Turbo V2.5 - no issue (436d3f75-08ab-4ef0-81d8-81d127eaaf4a) Eleven Turbo V2.5 - very slight cutoff of “Hi” (ef486666-b87a-4823-a77a-68f96bd0d875) Eleven Flash V2.5 - “Hi” was cut off (91d7954a-47a3-4b0f-8554-b6119c8958e1) Eleven Flash V2.5 - “Hi” was cut off (5c40d129-18e9-4359-baa4-0480781594fb) Eleven Multilingual V2 - no issue (20a8745d-36fd-4d7c-b948-ba0efdff5e0b) Eleven Multilingual V2 - slight interruption of word “Hi” (eae16a4e-62d7-44bc-90be-6b176e96aac9) Eleven Multilingual V2 - slight interruption (distortion) of word “Hi” (a04d5516-8b42-43df-9835-66e8a6397dd9) Eleven Turbo V2 - no issue (dee9e0b5-54a0-4ba4-a31a-3d666b619ff3) Eleven Turbo V2 - no issue (de0df303-bbee-48ab-bad7-93762172773b)

@MS for now you can use

eleven_turbo_v2

its better compared to all of the other models, I will share the flash model feedback with the 11labs team. Also I will update you with their response.