OpenAI just rolled out an upgrade to ChatGPT’s Advanced Voice mode. The voices now sound more natural with subtler intonation, realistic cadence including pauses and emphasis, plus better expressiveness for emotions like empathy and sarcasm.
The update also makes translating languages smoother. Just ask ChatGPT to interpret, and it will keep translating the conversation until you say stop or switch languages.
The feature is live for all paid ChatGPT users worldwide on every platform. OpenAI warns there might be minor dips in audio quality, like unexpected tone or pitch changes. 
The update doesn’t fix existing voice mode bugs, such as hallucination-related issues that cause unintended sounds, gibberish, or random background music.
OpenAI notes:
ChatGPT’s voices now sound more natural and fluid, with “subtler intonation,” “realistic cadence” (including pauses and emphasis), and more “on-point expressiveness” for emotions like empathy and sarcasm.
Voice mode now also allows users to translate languages more easily. Ask ChatGPT to interpret, and it will continue translating the conversation until you tell it to stop or switch to another language.
There may be minor dips in audio quality, including “unexpected variations in tone and pitch.”
This update doesn’t fix voice mode’s occasional hallucinations-related bugs, such as unintended sounds, gibberish, or background music.
