OpenAI has improved ChatGPT’s voice conversation feature and added two voices designed specifically for Korean speakers. The new voice model makes Korean intonation and expressions more natural while helping conversations continue smoothly even when the AI is searching for information.
OpenAI announced on the 16th that it had updated ChatGPT Voice using its GPT-Live-1.1 voice model.
The update introduces two new voices for Korean users: Hanul, a male voice, and Haetsal, a female voice. Users can select their preferred voice when talking with ChatGPT.
GPT-Live-1.1 improves intonation and vocal expression in Korean and several other languages. The enhancements are designed to make conversations more natural when users organize their thoughts aloud, ask questions or practice speaking foreign languages.
The way ChatGPT communicates while searching for information has also been improved. Previously, it repeatedly used phrases such as “Let me check” while preparing an answer. Following the update, it uses a wider variety of context-appropriate expressions to keep the conversation flowing.
ChatGPT Voice uses a “full duplex” architecture that allows it to listen and speak simultaneously. Users can ask follow-up questions while the AI is responding, and if they pause briefly, the AI can wait rather than immediately interrupting the conversation.
OpenAI introduced full duplex voice conversation technology when it launched GPT-Live in July. The latest update builds on that architecture, further improving multilingual intonation, vocal expression and conversational behavior during information searches.
The update is rolling out gradually to ChatGPT users worldwide, including in South Korea. It is available to free users as well. Users can access the feature by tapping the voice icon in ChatGPT.