I talked to Google Gemini Live for an hour; the lag was never the problem

Published Sep 22, 2026, 11:15 AM EDT Nimrod Aldea covers Android, wearables, and Google's AI tools for Android Police. He has spent more than 25 years reviewing consumer technology, with bylines at Cracked and Geek Universe, which makes him older than most of the things he writes about. He is most interested in the point where a gadget stops being a product and becomes part of somebody's ordinary life. He lives in Amsterdam, plays cozy games with more commitment than they strictly require, and owns physical media he can no longer play. Less than five years ago, the idea of having a conversation with a computer was relegated to science fiction. Now AI chats are everywhere, and we are completely OK with talking to a large language model about our job, our taxes, our vacation plans, and, for a large portion of Gen Z, sharing our emotions and asking for life advice. However, I was never able to shake the feeling that these conversations with computers never feel like a real conversation. At best, they feel like talking to a phone representative or someone from tech support. We turn to them when we have a problem, and they help us through it. The tech that's supposed to fix all of this No more speech-to-text-to-speech relay race Credit: Google OpenAI released GPT-Live on July 8, 2026, rolling it out globally in ChatGPT to replace the older turn-based Advanced Voice Mode. This was presented as a more natural way to talk to ChatGPT: something that feels much more like having a normal conversation with a real person. This is an entirely different technology than the voice features we've had until now, which listens to your speech, converts it into text, runs that text through an LLM, then sends the answer through a text-to-speech program that reads it back in a natural-sounding voice. GPT-Live, and Gemini Live right behind it, skip that whole relay race. Instead of translating your voice to text and back to voice, the model works with raw audio the entire way through (what it hears and what it says) inside the same system, with nothing translated in between. That sounds like a small plumbing detail, but it's the whole story. Cutting out the text step lets these models respond in a fraction of a second instead of the pause we've learned to expect, and it lets them hear things text can never carry: tone, hesitation, whether you're annoyed or joking. It's also why you can talk over it, or it can talk over you, without the exchange freezing while some hidden system decides whose turn it is. On September 15, Google released its version of this naturally flowing voice chat: Gemini Live. I sat down and really had a chat with it. I have talked to GPT Live a few times, but those were regular inquiries, mostly about places I can go to eat and interesting spots to visit while on vacation. Though it was noticeably faster than the regular chat, I didn't feel too much of a change from the previous version. So I actually put it to the test A walk built entirely around one long conversation Credit: Nimrod Aldea I put on my earbuds and went on a very long walk, entirely dedicated to having a deep conversation with Gemini Live. I was hoping to be surprised by how natural the conversation felt. Instead, I came out with a deeper appreciation for every human I've ever talked to. Even the boring ones. The conversation started with me trying to test the boundaries of what this voice model can do, and this part left me deeply impressed. I spent the first 15 minutes or so trying to stump Gemini and see when it stopped understanding me. The only one left stumped was me. It kept up across three languages and caught a joke I didn't explain The demo worked exactly as advertised Credit: Google I spoke, it spoke back. I spoke faster, it answered faster. Then I switched to a different language, and it switched along with me. Even switching between languages several times during the same sentence didn't stump it. The most impressive moment happened when I asked it what "T-O-P-G-3-3-K" spelled out, and it immediately came back with, "You are spelling the word Top Geek, but using a 3 to represent a reversed E." I don't know about you, but to me that points to either abstract thinking or a solid understanding of the internet concept of leetspeak. Probably the latter. At this point, I stopped trying to stump it and gave in to the conversation to see how natural it felt to speak to this advanced AI. I spent the next 45 minutes to an hour walking along a riverbank, telling Gemini about everything going on in my life and consciously trying to steer things toward a more personal, human-like conversation. We talked about my job, my family, and my plans for the future. Although it felt fast and responsive, at no point did it feel like talking to another human being. Not because the responses sounded robotic, or because the LLM couldn't understand what I was talking about. Gemini worked flawlessly throughout the conversation, exactly as Google planned. What Gemini couldn't replicate, because it was never built to replicate it, is human connection. Every AI chatbot is built for the same purpose: not to replicate a human, but to replicate an official representative. Someone who is there to answer your questions, not to listen. When we call someone on the phone, whether it's a parent, a friend, or a loved one, we are not calling them to get information. Sure, we exchange information, but the main purpose of that call is to hear their voice, see how they are doing, and let them know how we are. It's a connection, not a phone call. Gemini does not have a personal life, anecdotes to share, or any reason to care about us. I'm sure I'm not telling you something you don't already know, but somehow talking naturally to an LLM amplifies the feeling that there is no one on the other side of the line. It might flow like a phone call, but it doesn't feel like one. The kinks are gone. The conversation still isn't there Credit: Google None of this is a knock on the technology. Given the choice, I'd take this version of AI voice chat over the old one without a second thought. Google has quietly solved almost every frustration I've ever had about talking to an AI out loud. I went in assuming the awkwardness was a wiring problem, that talking to an AI felt strange because it worked like a walkie-talkie: one full sentence at a time, waiting your turn before you could speak again. Fix that, I figured, and the rest would follow. But the kinks are gone, and the feeling remains. What's on the other end isn't a person who forgot to show up. It's a remarkably smart echo, one that can answer anything you throw at it but has nothing of its own to bring to the conversation. Answering questions, no matter how fast or how well, was never the same thing as having a conversation.