A demo of ChatGPT Advanced Voice was given during the GPT-4o launch several months ago, however Open AI kept pushing back the release because of some security concerns. The ‘Sky’ voice then caused a stir since it bore a strong resemblance to Scarlett Johansson. At last, Open AI is now enabling Advanced Voice for all ChatGPT Plus and Team users, after a five-month wait. According to Open AI, the rollout will be finished this week.

If you haven’t heard, ChatGPT’s normal voice chat is far inferior than Advanced Voice, which is a great improvement. Using the GPT-4o model’s multimodal capabilities, Advanced Voice provides a realistic, free-flowing dialogue that accommodates interruptions.

This week, all Plus and Team users on the ChatGPT app will have access to Advanced Voice.

While you waited, we improved the accents, added five additional voices, added memory, and added personalized instructions.

More than 50 languages are supported for “Sorry I’m late” speech. tweeted on September 24, 2024 (@OpenAI) at tweet.com/APOqqhXtDg

Though ChatGPT‘s Advanced Voice and Google’s Gemini Live may sound similar, there is one important difference. Gemini Live extracts responses from an LLM and responds using TTS/STT (text-to-speech) engines; however, ChatGPT Advanced Voice provides audio input/output directly. Gemini Live does not provide a fully multimodal experience, but it does support interrupts.

However, based on my limited testing, it appears that several multimodal capabilities have been removed from ChatGPT Advanced Voice. Open AI demonstrated throughout the demo its ability to sing for you, recognize emotions and moods from your speech, recognize various noises, mimic accents, and much more. However, now, Advanced Voice indicates that it can’t distinguish the speech. Additionally, camera input is not yet supported.

It appears that Open AI disabled a few functionalities in order to prevent awkward chats using ChatGPT. Still, are you looking forward to using ChatGPT Advanced Voice? Tell us in the comments section below.