GPT-Live: OpenAI Launches Real-Time Full-Duplex Voice Conversations

These metrics underscore OpenAI’s commitment to safety, ensuring that the model’s lifelike conversational capabilities cannot easily be weaponized or lead to psychological harm.

Consumer Features, Global Rollout, and Limitations of GPT-Live

Starting today, GPT-Live is rolling out globally, completely redefining the default voice interface across the ChatGPT ecosystem. Paid subscribers on the Go, Plus, and Pro tiers gain immediate access to GPT-Live-1, while free tier users worldwide can experience GPT-Live-1 mini.

The updated experience introduces several consumer-facing enhancements:

• Nine Remastered Voices: OpenAI has completely remastered its suite of nine conversational voices—Arbor, Breeze, Cove, Ember, Juniper, Maple, Sol, Spruce, and Vale—specifically optimized for the nuance, tone, and pacing of the full-duplex engine.
• Rich Visual Integration: During spoken conversations on mobile or web browsers, ChatGPT will dynamically display beautiful visual cards for real-time data like weather, stock prices, and sports scores.
• Multimodal Alignment: Spoken responses stream side-by-side with generated text, allowing users to share text and images within the same