OpenAI rolls out GPT-Live-1 voice models

OpenAI is launching its new GPT-Live-1 voice models to all ChatGPT users. The update introduces continuous interaction, allowing the AI to listen and speak simultaneously. This architecture enables real-time translation and allows the model to delegate complex reasoning tasks to GPT-5.5 while maintaining a conversation.

The voice now incorporates filler words like “ums” and “ers” to mimic human speech. Users can also interrupt the AI with reduced lag.