OpenAI launches GPT-Live-1 voice models, full version requires paid subscription
OpenAI just launched GPT-Live-1 and GPT-Live-1 mini, voice models that let you talk to ChatGPT more naturally, even overlapping speech like a real conversation.
The mini version will replace the old Advanced Voice Mode for everyone, but if you want the full-featured GPT-Live-1, you'll need a paid subscription.
Models team with GPT-5.5 for visuals
These models can team up with text systems like GPT-5.5 during chats, so you get smarter answers and even visual responses.
They're built to handle long talks and can pause to "listen" before replying.
Still, during a demo the assistant had a heavy American accent and unnatural-sounding Hindi with a bookish tone, despite claims of language optimization.
And while OpenAI says these aren't meant as AI buddies (just hands-free helpers), they're rolling out alongside big competition from Apple, Amazon, and startups like Sesame.