Used ChatGPT's new voice mode to translate the World Cup in real time -- here's what happened
OpenAI’s new GPT-Live (including GPT-Live-1 and GPT-Live-1 mini) is presented as a major shift from prior ChatGPT Voice experiences, particularly for live, continuous audio. A reporter tested the feature during the 2026 World Cup knockout rounds by placing a phone near laptop speakers and letting ChatGPT listen to a Telemundo Spanish stream. Instead of waiting for pauses or cutting off when announcers spoke over each other, the system delivered a running English translation over the broadcast audio. The article attributes this to full-duplex architecture, describing simultaneous listening and speaking, plus a “decoupled delegation” approach for more intensive reasoning. It also says GPT-Live can display dynamic visual cards—such as dates, maps, and sports stats—while translating. The piece describes safety efforts by OpenAI and provides examples of prompt wording intended to trigger the simultaneous translation behavior.





