In this mode, voices sound like real people, as opposed to robotic voices.
OpenAI surprised everyone earlier this year when it unveiled an updated voice mode for its latest version of ChatGPT. Unlike the robotic voices of digital assistants like Alexa or Siri that people are used to, ChatGPT’s improved voice mode sounds like a real person. It plays in real time, pauses at the right time, can hear laughter when someone is making a joke, and can judge the emotional state of the person speaking by the tone of their voice.
As of July 30th, the latest version of the chatbot, an extended voice mode that works with ChatGPT-4o, has also become available to paying users. The extended voice mode is distributed among the group of subscribers of the “Plus” mode and is planned to be available to all “Plus” users from autumn.
According to OpenAI, AI voice capabilities have been tested in recent months by “testers who speak 45 languages and live in 29 different geographies.”
For security reasons, the voice mode cannot use any of the four types of voices created in conjunction with voice actors. AI will also block certain requests aimed at creating music or other copyrighted audio material.
The enhanced voice mode differs from the previously shown version in one important way: users won’t be able to use Scarlett Johansson’s voice, which OpenAI demonstrated in May. The company said the voice was actually created using another actor and that its use was stopped “out of respect” after the actress complained.
The launch of ChatGPT’s improved voice mode comes after OpenAI announced last week that it was testing an AI-powered search engine. Eventually, the OpenAI search engine could become a serious competitor to Google, the leader in online search.
