ChatGPT升级版语音模式更懂得适时闭嘴

OpenAI正在全面革新ChatGPT的语音模式，采用的新模型据称更像是“与另一个人交谈”。全新的GPT-Live-1旨在减少对你的打断，如果你在对话中途停顿，它还会等待你继续发言。

在一次新闻发布会上，OpenAI研究负责人Kundan Kumar称GPT-Live-1是公司迄今为止“最聪明的语音模型”。当需要进行推理或搜索网络时，它会自动将你的查询转交给最佳文本模型（如GPT-5.5），从而能更快地从研究你询问的主题过渡到讲述其发现。升级后的模型还会在谈论天气、股票和体育时，补充由AI生成的视觉效果，显示体育比分或本周天气预报等相关信息。

ChatGPT此前依赖较旧的、基于轮次的语音模型，该模型并非总能给出准确答案，有时也难以维持自然的对话流。“这是一个全双工模型，”OpenAI产品负责人Atty Eleti在发布会上表示，“它真正意味着可以同时说话和倾听……从模型侧来看，它可以连续且同步地处理输入流并产生输出流。”

这一设置为ChatGPT语音模式开启了新功能，例如实时翻译。现在，ChatGPT可以在你说话的同时进行翻译，而无需等到你停止说话。你现在还可以让ChatGPT语音在未被呼叫时保持安静，OpenAI表示这在以前是不可能做到的。它还会用“嗯嗯”、“是的”和“明白了”等短语来确认自己正在听。

OpenAI指出，已添加内置安全防护措施，引导模型远离有害回复，或在“更高风险”情况下完全终止聊天。（OpenAI目前正面临一系列诉讼，指控ChatGPT助长妄想、并伤害了用户心理健康。）该模型还经过训练，可在涉及自残的对话中提供“专家审核的危机求助支持”，同时也会为青少年提供“适龄”回复。

GPT-Live-1正在iOS、Android及网页端推出。它将为Go、Plus和Pro订阅用户驱动ChatGPT语音功能，而免费用户则将默认使用更小、更高效的GPT-Live-1 mini模型。

OpenAI is overhauling ChatGPT’s voice mode with a new model that it says is more like “talking to another person.” The new GPT-Live-1 is designed to interrupt you less and will also wait for you to continue speaking if you pause mid-conversation.

During a press briefing, OpenAI research lead Kundan Kumar called GPT-Live-1 the company’s “smartest voice model” yet. It will automatically pass your queries to its best text models, like GPT-5.5, when it needs to reason or search the web, allowing it to more quickly transition from researching the topic you’ve asked about to talking about its findings. The upgraded model will also supplement conversations about the weather, stocks, and sports with AI-generated visuals showing relevant information like sports scores or this week’s forecast.

ChatGPT previously relied on an older, turn-based voice model, which didn’t always produce accurate answers and sometimes struggled to maintain a natural conversational flow. “This is a full duplex model,” OpenAI product lead Atty Eleti said during the briefing. “What it really means is that it can speak and listen at the same time. … From the model side, it can process the stream of inputs and produce the stream of output continuously and simultaneously.”

This setup opens up new features for ChatGPT’s voice mode, such as real-time translation. Now, instead of waiting until you stop talking to translate your speech, ChatGPT can translate while you talk. You can also now ask ChatGPT Voice to stop talking until it’s called on, which OpenAI says wasn’t possible before. It will acknowledge that it’s listening with phrases like “mhmm” or “yeah” and “got it,” too.

OpenAI notes that it has added built-in safeguards that steer the model away from harmful responses or end chats completely in “higher-risk” situations. (OpenAI is currently facing a string of lawsuits alleging ChatGPT fueled delusions and harmed the mental health of its users.) The model is trained to offer “expert-vetted crisis helpline support” in conversations about self-harm as well, and is also designed to provide “age-appropriate” responses for teens.

GPT-Live-1 is rolling out across iOS, Android, and on the web. It will power ChatGPT Voice for Go, Plus, and Pro subscribers, while a smaller, more efficient GPT-Live-1 mini model will be the default for free users.