There was multiple sequential model outputs. For some reason, last model output has not been voice. At the same time, user said "OK", Can it be the potential cause? Also, for some reason, there is not logs related to user input (no transcriber logs, etc.). Here is the logs fragment:

01:36:27:311
[LOG]
Voice input: Your total is $19.77.

01:36:27:311
[LOG]
Model output: Your total is $19.77. <flush/>

01:36:27:604
[LOG]
Model output: Please wait a moment while I submit the order and get the confirmation. <flush/>

01:36:27:605
[LOG]
Voice input: Please wait a moment while I submit the order and get the confirmation.

01:36:27:657
[LOG]
Model output: You'll receive the order summary on your phone shortly. <flush/>

01:36:27:658
[LOG]
Voice input: You'll receive the order summary on your phone shortly.

01:37:08:080
[LOG]
Model output: Thank you, AC. Your order for a Falafel Burger, a Snapple, and a side of fries has been placed. Your total is $19.77. You'll receive the order summary on your phone shortly. If you have any further questions or need assistance, feel free to ask. Enjoy your meal!... <flush/>

01:38:13:544
[LOG]
Transcriber output: Hi.

01:38:13:746
[CHECKPOINT]

The following model output has not been voiced:

01:37:08:080
[LOG]
Model output: Thank you, AC. Your order for a Falafel Burger, a Snapple, and a side of fries has been placed. Your total is $19.77. You'll receive the order summary on your phone shortly. If you have any further questions or need assistance, feel free to ask. Enjoy your meal!... <flush/>

Call ID: https://dashboard.vapi.ai/calls/caf8d426-5feb-43e0-83e1-cd04dd544921

@Moody Reaper It seems like there might have been an issue with the Text-to-Speech (TTS) process for the final model output. The user saying "OK" should not typically interfere with the TTS process, as the system is designed to handle simultaneous operations. However, the absence of transcriber logs for the user's input around that time could indicate a problem with the voice input capture or logging system. To address this issue, I would recommend checking the TTS service status to ensure it's operational and reviewing the system logs to identify any errors that occurred during the time of the incident. If the problem persists or if there is no clear indication of what went wrong, please reach out to support for further assistance.

It was an issue from the 11labs side, that particular voice input the WebSocket wasn't able to connect. ElevenLabs (Websocket #25) Not Open. Punting 277...

@Sahil, thank you! Is it possible to do something with that? In some cases, these "unvoiced" model output pieces are very critical from business perspective (user flow)

For now, I could suggest you to use Cartesia because lately, I have been seeing a lot of performance degradation in 11labs.