Currently, when using “talk while waiting”, the moment the tool returns a result the system immediately cuts the current audio, generates the next TTS response, and sends it to the user.
This feels unnatural because it interrupts the sentence and creates a short, awkward moment of silence.
Ideally, when the tool result is received, the system should allow the current spoken sentence to finish while generating the next TTS response in the background. Once the sentence completes, it can transition into the new response. This would make the interaction feel much more natural and conversational.