Realtime Voice AI

Realtime is a narrower claim than voice. A tool belongs here if it is built for live, interruptible speech — streaming recognition, sub-second response, and the ability to stop talking the moment someone else starts. That is an engineering problem as much as a model problem, which is why the field is smaller than the general voice category and why the products are more often sold as infrastructure than as finished apps. If you are building your own agent rather than buying one, this is the category to start in.