Many people want a voice AI that works like a real interpreter. OpenAI moved closer to that idea on May 7, 2026, when it announced GPT-Realtime-Translate. According to OpenAI, this new model can translate speech from more than 70 input languages into 13 output languages, and it can keep up with the speaker instead of waiting for a full recording to finish. OpenAI’s audio documentation also explains that live translation should use a dedicated realtime session, so translated speech can begin as audio arrives. (openai.com)
So, can OpenAI’s new voice AI interpret while people are still talking? In the technical sense, yes. The new model is designed for streaming speech-to-speech translation, and OpenAI says it can return translated audio and transcript updates while the original audio is still coming in. That is much closer to human-style interpreting than older systems that first make a full transcript and only then produce a translation. However, there is an important point: OpenAI introduced this feature in its API for developers. In OpenAI’s public ChatGPT Voice FAQ, ChatGPT voice is described as a spoken conversation tool for logged-in users, but the help pages do not present this new live translator as a standard, built-in ChatGPT voice mode for everyone. (developers.openai.com)
For English learners, this is exciting. A future app could let you speak Japanese, hear quick English output, and check the transcript later. But it is not magic. OpenAI clearly warns that voice conversations may make mistakes, and its help pages say voice input can sometimes detect the wrong language unless you set your main language in the app settings. So the best answer is: yes, OpenAI’s newest voice technology can already do live interpreting-like translation, but it is still a tool that needs careful use. For travel, study, and casual conversation, it could be very helpful. For medical, legal, or business situations, people should still double-check important meaning. (help.openai.com)










