Character Calls turned character chat into a live turn-taking surface

From voice output to calls

Character.AI announced Character Calls on June 27, 2024 as a two-way voice conversation available on its app and web. Its release described tap-to-start calls, interruption, different voices and the ability to move between text and speaking. These mechanics matter because a companion interaction is no longer bounded by composing and reading a message: the system has to manage tempo, microphone turns, interruptions, and a handoff back to text.

The earlier March 2024 Character Voice update provides the immediate precursor. It described voices assigned to one-to-one chats, a community voice library, user-created audio samples, and public discovery of shared voices. Calls therefore did not introduce voice from nothing; they converted an existing character attribute into a synchronous conversational mode. The release called the experience free at launch, a historical availability detail rather than a present pricing guarantee.

What the mechanism changes

Evidence: the two first-party announcements document the sequence and stated controls. Inference: interruption is especially consequential in this format. It lets users repair a misunderstood turn quickly, but it also makes the exchange feel less like an asynchronous tool and more like an interlocutor. That can improve flow without establishing that the character understands intent more deeply. The announcements do not disclose the speech stack, retention behavior, or real-world latency distribution, so those should not be inferred from ‘reduced latency.’

Evidence limit: Feature claims and availability are company statements; no hands-on test or current entitlement review was performed.