You pause to remember a number. Say “yeah” while someone explains. Change your mind halfway through an order. Those moments are part of a conversation.
Today, we’re introducing Live 3, Fish Audio’s full-duplex voice model. It keeps listening while it speaks, so a pause, a quick acknowledgment, or a change of mind can shape what happens next.
A quiet “mm-hmm.” Space to finish a thought. Room to jump in.
Live 3 is built to give people room to speak naturally. Try it for yourself at fish.audio/agents.
Learning the rhythm of conversation
Speed and conversational timing are not the same problem. An agent can start a reply quickly and still speak at the wrong moment. A pause after “What time do you open?” might mean you’re done. A pause halfway through a phone number means there is more to come.
Most voice agents coordinate speech recognition, reasoning, speech generation, and turn-taking across separate parts of a pipeline. Getting those parts to work together naturally takes care. Live 3 brings conversational timing into the speech model itself.
We trained it on real human conversations so it learns when to acknowledge, wait, respond, or stop. Those behaviors were learned from people, not written as a list of rules.
“The fix so far has been more rules and more thresholds,” said Rissa Cao, Fish Audio’s co-founder and CEO. “We think a model should learn this from real conversations, the same way language models learn grammar.”
Acknowledging without taking over
Read out a number in chunks: “415… 555… 0132.” Live 3 might say “mm-hmm” or “okay, 415” between them and wait for the rest.
That small acknowledgment does two things at once. It tells you the digits were heard, and it leaves the floor open. You can take a breath, check the next few numbers, and carry on.
Live 3 can decide to respond in as little as 186 milliseconds. Short reactions such as “mm-hmm,” “okay,” and quick confirmations keep the exchange moving while the full answer is being prepared.
Knowing when you want to jump in
Not every sound is an interruption. If an agent is explaining its services and opening hours, you might say “yeah” or “mm-hmm” to show you’re following along. Live 3 keeps going.
Then you say, “Wait, actually, I just need Thursday.” It stops and makes room for your new request.
That distinction changes the feel of the whole call. You can acknowledge what you hear without accidentally cutting it off, and change direction without waiting for a long answer to finish.
Giving you room to change your mind
“I’d like two pizzas… no, three. Three large. And… actually, make one of them a margherita.”
Live 3 waits through each correction instead of jumping in with the wrong item. You are still deciding what you want. It leaves room for that thought to finish.
Support calls, bookings, and orders are full of these moments. People hesitate, remember another detail, and revise a request halfway through saying it. An agent that follows that rhythm lets the conversation move forward without making the caller start over.
Hear the difference
A voice can sound natural on its own. The real test is what happens when someone talks back.
Listen to three excerpts from real test calls, with the original pauses and overlaps.
Room to finish your order The caller adds details with pauses. Live 3 acknowledges along the way, then reads the item back.
Jump in when you’re ready The caller cuts into the menu list with an order. Live 3 stops, listens, and confirms it.
A quick “yeah” keeps the conversation going The caller says “mm-hmm” and “yeah” while Live 3 reads the order back. It carries on.
Live 3 listens throughout the exchange: acknowledging without taking over, leaving space when more is coming, and yielding when you cut in. You bring the instructions, tools, and workflows that make the agent yours. Live 3 brings the rhythm of conversation.
Live 3 speaks English today. It is available on the Fish Agent Platform and through our API, and enterprises can deploy it on-premises.
Pause. Think. Change your mind. Keep talking.
Try Live 3 at fish.audio/agents →
