New voice AI starts looking up answers while you are still talking

A company called Phonic has released a voice AI model, Phonic-1, that starts working on an answer before you finish your sentence. It announced the model on October 29, 2026. The goal is to cut the awkward pause between asking a question and hearing a reply, a pause many people know from phone assistants.

What changed

Many voice AI systems turn speech into text first. Then the AI writes a reply and reads it aloud. Tasks like searching for information or confirming a reservation add more waiting. Tone of voice and the silences between words also get lost when speech becomes text.

Phonic-1 works with speech directly, a method called Speech-to-Speech. It can start needed tasks while the other person is still talking. If you ask about travel, it can begin searching while you speak.

The numbers so far

Phonic published a comparison demo on its website. In it, Phonic-1 used five outside tools while the speaker was still talking. It answered 1.1 seconds after the speaker stopped. OpenAI's GPT-Live, the comparison tool, waited until the speaker finished before using a tool. It took 15.7 seconds to reply.

These figures come from Phonic's own demo, so they show a best case chosen by the company, not an independent speed test.

How natural does it sound

Phonic says Phonic-1 can pick up the tone of a person's voice and judge when to speak, even in noisy places. A third-party group, Podonos, ran a voice comparison. Listeners preferred Phonic-1's responses over GPT-Live-1's in 61.5% of cases.

Against a real human voice, people picked the human 57.0% of the time. That suggests Phonic-1 still has room to improve, and Phonic acknowledges it.

Who it is for

Phonic plans to bring voice AI to customer support, job interviews and logistics. Businesses that handle many calls are the clearest audience. Phonic-1 was trained on more than 50 languages. It can keep the same voice if the conversation switches languages midway.

• Phonic-1 is available now, according to the report.
• The material does not give pricing.
• It does not say how businesses can sign up or what they need to set it up.

What to watch

For anyone weighing voice AI for customer calls, shorter pauses could make conversations feel more natural. The speed claims still need testing outside Phonic's demo. Real calls involve messy questions, bad connections and slow outside systems, and those are the conditions that matter.