This week, all ChatGPT users finally gain access to the new GPT-Live voice mode, which introduces simultaneous two-way conversation. The model, announced in previous weeks, is now available for both free and paid users, albeit with slightly different versions: GPT-Live-1 mini for free users and GPT-Live-1 for subscribers. OpenAI has acknowledged some initial bugs, but the rollout is proceeding smoothly for most users.
How to check availability and differences between versions
The easiest way to know if you have the update is to look for the Live label at the top of voice chats on mobile or behind the ChatGPT dropdown on the web. Tapping or clicking these labels allows you to revert to the old voice models, at least for now. On mobile, you can also check via the settings menu: tap the menu button (top left), then the settings cog (Android) or profile avatar (iOS), then Voice and Model. On the web, click your profile avatar (bottom left), then Settings, Voice, and Model.
Sponsored Protocol
Duplex functionality revolutionizes voice interaction
The main novelty of GPT-Live is its duplex capability, allowing you to speak while ChatGPT is responding, without waiting for it to finish. Try interrupting it mid-sentence: the system immediately adapts, reformulating its response in real time. This fluidity brings the experience closer to a real human conversation, with pauses and variations in speech rhythm. Initial tests show the system works as advertised, with only rare glitches.
Real-time translations and visual cards expand possibilities
Another notable feature is simultaneous translation: you can speak phrases in English and ChatGPT translates them into another language on the fly, without hesitation. Although not yet ideal for language learning due to some imperfections in the pronunciation of foreign languages, it represents a significant step forward. Additionally, OpenAI has added visual cards to voice mode: when asking about sports scores, weather forecasts, or directions, graphics and maps appear on screen. Simply click the sliders icon at the top right to change voice and intelligence level (Instant, Medium, or High).
Sponsored Protocol
Known bugs regarding memory and accents, but rollout continues
Not everything is perfect: OpenAI has reported that ChatGPT memory may be unreliable in some voice sessions with GPT-Live. The company is actively investigating and will follow up with updates. On Reddit, engineer Atty Eleti has answered user questions, also confirming issues with English-accented pronunciation for foreign languages. However, most users have not encountered major issues. In the broader AI landscape, while OpenAI advances with GPT-Live, other models like Gemini struggle in app development tests, as highlighted in a recent article on Google updates Android Bench. For more on the implications of this technology, see the Wikipedia page on ChatGPT.
Sponsored Protocol
