1. Assistant is saying the following text weirdly. Call ID c77281a5-8289-4be1-9d28-516e92fc40c7 Input text "XXXX is available on March, four, Twenty Twenty five, from three in the afternoon to three forty in the afternoon." Assistant is saying "March 4 tea from 3 to 3 40 in the afternoon." 2. Input text "call nine one one". Assistant gives a pause between 91 and 1.

HeyDhina Thank you for creating the ticket! A member of our team will reach out to you shortly.

@Dhina For the first issue, it seems like there might be a problem with the text-to-speech (TTS) interpretation of the date and time format. To address this, we can adjust the input text to be more TTS-friendly. Try rephrasing the input text to: "XXXX is available on March fourth, two thousand twenty-five, from three PM to three forty PM." For the second issue, the pause between "91" and "1" could be due to the way the numbers are being processed. To ensure the numbers are spoken continuously without a pause, you can try inputting them as: "Call nine-one-one." If these adjustments do not resolve the issues, please ask support for more info on this.

Call ID - d8312efb-81f6-4bc5-ba9f-b8cb3291fb40 XXX is available on March, third, Twenty Twenty five, from three forty in the afternoon to four forty in the afternoon. Bot is saying March 3 Road

I also ran into this a few times yesterday. I can't hear your call ids of course, but I noticed when the agent was repeating back phone numbers or zip codes, sometimes it would start incomprehensively muttering for a few seconds in between numbers, and then continue on doing the same for a few seconds after it finished reading back the numbers.

@Dhina It is because of the assistant voice formatting plan, you have to change the default voice.chunkPlan.formatPlan config according to your requirements, do give it a read https://docs.vapi.ai/documentation/core-concepts/assistants/voice-formatting-plan

Can you please provide me an example of how to disable this option

🔵 13:52:41:458 Voice Input Formatted: "from three forty to four forty in the afternoon. - March 4th,", Original: "from three forty to four forty in the afternoon. - March 4th," 🔵 13:52:41:586 Voice Input Formatted: "from three to three forty in the afternoon. - March 4th,", Original: "from three to three forty in the afternoon. - March 4th," 🔵 13:53:04:686 Voice Input Formatted: "Sure thing. Here are the available slots with doctor Dueno in March:. - March 3rd,", Original: "Sure thing. Here are the available slots with doctor Dueno in March: - March 3rd," 🔵 13:53:05:235 Voice Input Formatted: "from three forty to four forty in the afternoon. - March 4th,", Original: "from three forty to four forty in the afternoon. - March 4th," 🔵 13:53:06:020 Voice Input Formatted: " from four twenty to four forty in the afternoon. Which date and time works for you?", Original: " from four twenty to four forty in the afternoon. Which date and time works for you?" @Dhina As you can see it is coming from transcriber side where as the input to voice providers are correct, so you can ignore the previous message. The solution here is to use modelOutputInMessages instead of transcribed output in messages. You can enabled it by making an UPDATE request to your assistant. UPDATE /assistant/:id

example_curl_request

curl -X PATCH https://api.vapi.ai/assistant/insert-your-assistant-id-here \
     -H "Authorization: Bearer insert-your-private-key-here" \
     -H "Content-Type: application/json" \
     -d '{
  "modelOutputInMessagesEnabled": true
}'

Thank you