OpenAI launches GPT-Live, making ChatGPT sound more human
What's the story
OpenAI has just launched a new voice model for its AI chatbot, ChatGPT. The company claims the new model, called GPT-Live, will make conversations with the chatbot sound more human-like. OpenAI CEO Sam Altman described the update as "magical" and "real," suggesting a significant leap in natural language processing technology.
Technological advancement
GPT-Live replaces advanced voice mode
The new voice model comes in two versions, GPT-Live-1 and GPT-Live-1 mini, which will replace the older advanced voice mode of ChatGPT. These models are designed to minimize interruptions during conversations while performing complex tasks in the background. This means you don't have to wait for ChatGPT to process information like translations or research before continuing your conversation.
Innovative design
Full-duplex architecture at the core of GPT-Live
At the heart of GPT-Live is a full-duplex architecture, which allows the AI to listen and speak at the same time. Unlike older voice systems that took turns responding, GPT-Live continuously processes information while talking. This lets it use human-like interjections such as "mhmm," "yeah," or "got it" during conversations, making interactions feel more natural and fluid.
Enhanced functionality
The new voice mode can also show visual cards
GPT-Live can also handle complex tasks like search and reasoning by delegating them to another system in the background, currently GPT-5.5. This was demonstrated when it translated a presentation into Hindi while the speaker was still talking. The new voice mode can also display visual cards on topics like weather, stocks, and sports while continuing to support search, memory, images, and file uploads.
User safety
OpenAI has also expanded safety testing for GPT-Live
OpenAI has also expanded safety testing for GPT-Live with audio-native and synthetic evaluations. These tests focused on self-harm, psychosis and mania, emotional reliance, violence, and sexual content. The company said the model comes with safeguards that can steer responses toward safer output or end a voice conversation in higher-risk cases. OpenAI has also added protections for teen users such as age-appropriate behavior and parental controls.
Rollout strategy
Availability and API access for developers
OpenAI has confirmed that GPT-Live is now available for all users on iOS, Android, and the web. The company also plans to make it available via API in the near future. Developers and enterprises can sign up for access notifications. However, at launch, GPT-Live doesn't support voice with video or screen sharing inside ChatGPT, these features will continue to be offered through older voice modes.