After halting the development of a new generation of synthetic voices due to complaints from the actress, OpenAI has reintroduced them in the app for some users.
ChatGPT has regained its voice, although technically, this artificial intelligence never actually lost it. OpenAI, the company behind this popular tool, has started rolling out the new generation of its advanced voice models to some users of the ChatGPT Plus subscription service. The goal is to make these incredibly lifelike voices available to all paying users by fall.
None of these voices sound like Samantha, the artificial intelligence portrayed by Scarlett Johansson in the film Her. Last May, when OpenAI announced this new generation of voices for the conversational AI ChatGPT, many noted the striking resemblance of one of the options, named Sky, to the actress. Johansson even considered filing a lawsuit against the company.
The application now offers four voices: Juniper, Breeze, Cove, and Ember, which have been developed from voice samples of actors and actresses who were compensated for their work. “ChatGPT cannot impersonate other individuals and will block any attempts to deviate from these preset voices,” the company assures.
The delay has not left ChatGPT voiceless during these past months. Mobile app users have been able to maintain voice conversations instead of using written text, if that was their preference. However, these were somewhat older synthetic voices that were not integrated into the company’s latest language model, GPT-4o.
This made the experience somewhat slower. The application had to use a tool to transcribe user requests, feed the results to the language model, and then use the response generated by the model in a voice synthesis tool.
Now, all these steps occur within the language model itself, which is capable of understanding spoken text and responding vocally in a native manner. This means that voice conversations with ChatGPT will be faster and more fluid.
The company has showcased the speed of the new model in a video, which closely resembles what one would expect in a conversation with a human being. The language model can add different intonations and also incorporates pauses and fillers that contribute to making everything sound very natural.
Not all ChatGPT users will have access. This new generation of voices will only be available to users of the ChatGPT Plus subscription service, which expands the limits of the free version of the tool and grants access to more advanced configuration options.
The deployment will also be gradual and initially only in English. A few ChatGPT Plus users will receive access this week, with availability gradually expanding over the coming months.
