Today, we're launching Mia & Leo: two AI personalities for revenue calls, powered by MetaVoice.
MetaVoice is a purpose-built duplex speech model. It keeps listening while it speaks, allowing Mia and Leo to handle interruptions, overlapping speech, corrections and background voices without turning the call into a walkie-talkie exchange.
Sounding human is not enough
Voice agents have become much better at sounding human. But sounding human is not the same as holding a human conversation.
Teams building real-world agents told us that, on their existing voice aegnt stacks, over 40% of people ended the call within the first 30 seconds. The conversation broke before the workflow had a chance to matter.
The failures often feel small: an awkward pause, a missed interruption, or the agent talking over the person. They correct a date, but the agent does not catch it. A voice in the background causes the agent to lose the thread.
Then the person gives up.
On a revenue call, every lost conversation has a number attached: a missed booking, a lead that goes cold, a payment not collected, a patient not re-engaged or a customer who calls a competitor instead.
Today's voice stacks are not built for conversation
Most voice stacks work like this: Speech-to-Text → LLM → Text-to-Speech.
Each component waits for the one before it. People do not. We interrupt, overlap, pause and correct ourselves halfway through a sentence.
Traditional voice stacks often miss this conversational timing, creating friction that causes people to disengage or end the call.
A model that keeps listening
MetaVoice is built around one simple idea: a voice agent should keep listening while it speaks.
Powered by MetaVoice, Mia and Leo treat each call as one continuous, two-way exchange. They can respond to interruptions, corrections, overlapping speech and background voices without losing the thread.
Built for production
A model is not production-ready just because the demo sounds good. Developers also need control over agent behavior, visibility into calls, failures they can debug and costs that work at scale.
MetaVoice is built for that. It is designed for developers building inbound and outbound voice agents for calls with measurable outcomes: booking new leads, collecting payments and re-engaging customers or patients.
From human-sounding speech to human conversation
We started with speech that sounds human. Mia and Leo show what comes next: AI personalities that can hold a human conversation.
Talk to Mia and Leo at metavoice.io.