Imagine a conversation where the other person stays silent for a long time before every answer. Even a good answer would feel awkward. On October 30, a Japanese company announced an update to a tabletop talking robot. It can now begin speaking the first part of a reply while the rest is still being prepared.

A program turns written words into sounds. This is called speech synthesis^[1]. Instead of waiting for the whole sentence, the robot can play completed pieces as they arrive. This is streaming playback^[2]. The company also improved how some versions generate replies using artificial intelligence^[3], but the change is not available in every version.

There are trade-offs: streaming cannot be combined with some saved-audio and interruption settings. Reply speed depends on the internet connection, and no independent timing comparison has been published. The robot has not suddenly learned to understand everything; its conversation timing has improved. For reference, the product is called Sota.