OpenAI добавила одновременную речь и слух в голосовой режим для переводов

OpenAI on July 8, 2026 introduced an updated voice mode that enables AI to listen and speak simultaneously in a single moment. According to the company, this is a key improvement for direct speech translation — essentially the primary use case for this mode.

How Simultaneous Speech Works

The main difference from previous voice interfaces is simultaneity. Previously, systems worked sequentially: they either listened to the user until the phrase was complete, or they responded. The user waited for the AI to finish speaking, then started again. This alternating mode was noticeable in every dialogue — pauses between exchanges, the need to wait for a complete response before asking the next question.

The new mode can listen and speak in parallel. This brings interaction with AI closer to a natural conversation between people, where participants can interrupt each other, ask clarifying questions mid-thought, and react quickly to what they hear. The system can now process new information from the user even while it is still formulating a response to the previous question.

• Key improvement: simultaneous listening and speaking mode
• Application: direct speech translation in real time
• Effect: more natural dialogues, minimal pauses between exchanges

Real-Time Translation

The main use case is live translation. A person speaks in one language, the system translates in parallel and delivers the response in another, without requiring the user to wait for the phrase to be completed. This is critical for business negotiations between foreign teams, international conferences, calls with partners from other countries, lectures for students of different nationalities.

Previously the process was slow: the system had to listen to the user's entire phrase, then wait for the system to translate, and finally receive the response. Human natural rhythm requires much faster exchange. The new mode reduces this delay, allowing the system to react in parallel, almost like two people in a conversation.

Voice Interfaces Approaching the Norm

The key challenge with voice systems has always been the artificiality of dialogue. People are accustomed to a certain pace of conversation, to natural interruptions, to the ability to change topics mid-thought. Machine interfaces required the user to adapt to their rhythm.

Simultaneous speech and listening is not a revolution, but an important step. The impact extends beyond translations: support bots, virtual assistants, conference systems, tutors for online learning — all will gain a tool for more natural interaction.

What This Means

The update shows the direction of voice interface development: they should feel like a conversation with a person, not an interaction with a machine. The closer the system gets to the human pace of communication, the more natural the interaction feels. This could be an important step toward voice interfaces that are indistinguishable from talking to a real person.