Hi team, I’m using Vapi Squads with multiple assistants and i'm running into a frustrating issue. There is a 3–4 second delay (sometimes up to 9-10 seconds) of dead silence whenever the first assistant performs a silent transfer to the next assistant in the squad. This happens before the next assistant starts speaking, even with short/optimized prompts. I've tested with different LLM models that are faster and it is the same amount of time for the delay, it feels like it isn't the llm or not a big part of it. Is there any way to fix this? I believe that this delay issue has to be related to the silent transfer (transferCall tool). Here are some call-id examples: a109ef72-ea15-4b06-a5e3-8ddd3b557dda Timestamp: 00:10 - The next assistant (scheduling assistant) begins response at 00:14 4659b59a-a91e-49d5-9d50-2c1aec1882c1 Timestamp: 00:08 The next assistant (scheduling assistant) begins response at 00:18 Happy to provide more details if helpful.

To help reduce the silent transfer delay in Vapi Squads, ensure your destination assistant is configured as follows:

json
{
  "firstMessage": "",
  "firstMessageMode": "assistant-speaks-first-with-model-generated-message"
}

Also, set all

assistantDestinations[*].message

properties to an empty string. This setup minimizes transition artifacts and helps the next assistant start speaking promptly after a transfer. [See Silent Transfers Guide](https://docs.vapi.ai/squads/silent-transfers#silent-transfers) Source: - [Silent Transfers Guide](https://docs.vapi.ai/squads/silent-transfers#silent-transfers)

yes i have the destination assistant's first message set to "Assistant speaks first with model generated message" and i also have empty string in the destination message, but the issue still persists.

Try increasing silenceTimeoutSeconds, enabling warm transfers, optimizing speech timing, ensuring correct use of the transferCall tool, and using audio caching to reduce the delay..