ChatGPT’s New Voice Models Can ‘Listen’ and ‘Talk’ at the Same Time

Source Domain: www.cnet.com

Here are the key points from the article about the updates to OpenAI’s AI voice technology:

Introduction of New Models: Two new voice models, GPT-Live-1 and a mini version, are being rolled out to all ChatGPT users.
Enhanced Conversational Ability: The voice models sound more human and are more conversational with the ability to provide responses while actively listening.
Continuous Interaction Framework: The new framework allows the AI to simultaneously receive and produce responses, facilitating live, simultaneous translation.
Task Delegation: The AI can offload complex tasks to frontier models like GPT-5.5 to answer questions in the background and seamlessly integrate the responses into ongoing conversations.
Visual Answers: Uses AI-generated visuals for answers like weather reports and sports scores.
Accessibility: The new voice models are available now for all users, including free and paid subscribers, on mobile and web platforms. Rollout might take a day or two.
Human-like Speech Features: The voice attempt to mimic human speech with filler words and allows for fewer lags and more natural interruptions.
Safety and Privacy: Audio data is stored for 30 days for context continuity, can be deleted, and the models have enhanced safety measures for potential dangers associated with anthropomorphized AI.