OpenAI has unveiled GPT‑Live, its new voice model, which completely changes the way we talk to ChatGPT. Now we can have much more natural and fluid conversations with the AI and get the feeling that we’re really talking to a person. The new system is so surprising that, from the very first use, it’s clear we’ve moved to a new level of interaction.
Natural conversations without pauses
The key to GPT‑Live lies in its full‑duplex communication architecture, capable of listening and speaking at the same time. This means we can talk to it while it responds, interrupt it, or let it give us a few seconds to think, without cuts or artificial silences. It even uses expressions like “mhmm” or “I see” to keep the rhythm of the conversation and show us that it’s still there.
During the presentation, GPT‑Live demonstrated, for example, its simultaneous translation capabilities, in which ChatGPT knew to wait until the original sentence had enough meaning and context to do the translation, instead of going phrase by phrase or, worse, word by word.
It was also striking how it’s able to interrupt while a person is practicing speaking English to make relevant corrections. The naturalness of the interaction, even its laughter at certain points, is simply spectacular.
Available now for all users
The new version is already available in ChatGPT on iOS, Android, and the web, with GPT‑Live‑1 for those of us on the Go, Plus, and Pro plans, and GPT‑Live‑1 Mini for those using the free version. In the next few days it will also be available in the API for developers.
With GPT‑Live, talking to ChatGPT changes completely. A much simpler and more fluid interaction that invites you to use it at any time of day.
