ChatGPT speech model sees major improvements, tone, breaks hard to distinguish from real people

In the latest update to theChatGPTBringing a major improvement, this brand new version supports voice manipulation and communication, with the help of a new text-to-speech model and a speech recognition system that allows users to engage in voice conversations and choose from five different voice styles among them. This feature is already available in the iOS and Android apps.

Some netizens are interested inChatGPTA simple test of the voice was performed and the results showed that theAIHas been able to carry on conversations in a realistic manner, even in Chinese conversations, showing a level of performance not dissimilar to that of a real person, including fluent responses and appropriate tone of voice.

Most impressively, in voice conversations, ChatGPT behaves very naturally in terms of sentence breaks and tone of voice, and even adds in some verbiage such as "uh", which makes for a very comfortable conversation experience that is completely free of the electronic characteristics of previous AI voice models.

Combined with ChatGPT's powerful text generation capabilities, it has become almost impossible to distinguish it from a real person.

OpenAIThe new speech technology, which can generate realistic synthesized speech from a few seconds of real speech, opens up new possibilities for many creative and accessible applications, said the company.

However, these features also pose some potential risks, for example, phone scammers may use this technology to increase their success rate and may even pose as the victim's friends, relatives or public figures to perpetrate the scam.

OpenAIstated that they will be putting in place strict controls and restrictions to ensure that the model is not widely abused. However, it is also a reminder to remain vigilant in the current era of highly advanced AI and internet technology so as not to fall victim to online scams.