(Bloomberg) -- OpenAI is rolling out a much-anticipated voice assistant for ChatGPT to a limited number of users after previously delaying the release to work through potential safety issues.
The artificial intelligence startup said it would make the voice feature available to a small group of paid ChatGPT Plus customers starting on Tuesday. The product will offer four preset voices, but won’t be able to impersonate how other people speak, according to the company.
OpenAI also said it has added new filters to ensure the software can spot and refuse some requests to generate music or other forms of copyrighted audio.The voice option was a centerpiece of OpenAI’s product launch event in May for GPT-4o, an updated version of its GPT-4 model that is better at handling text, audio and images in real time. The startup had introduced a more limited option for ChatGPT to talk back to users last year, but the new feature promised to be faster and pair with powerful image-recognition capabilities to turn the chatbot into a far more useful and dynamic conversational partner.
OpenAI previously said that it had intended to roll out the voice assistant to a small group of paid ChatGPT Plus subscribers in late June, but decided it needed another month to “reach our bar to launch.” The company said it was working on the software’s ability to spot and refuse to respond to certain types of content, and ensure it could effectively process requests from millions of users. “By launching gradually, we can closely monitor usage and continuously improve the model’s capabilities and safety based on real-world feedback,” the company said in a statement Tuesday.
In the statement, OpenAI reiterated its plan to roll out the voice feature to all of its paid ChatGPT Plus subscribers in the fall. The company also said it is still working on video and screen-sharing features that were demonstrated during the May event. Those features don’t yet have a launch date.
As a result, the voice feature will have a more limited list of capabilities to start. For example, the chatbot won’t be able to access a computer-vision feature that would let it offer spoken feedback on a person’s dance moves simply by using their smartphone’s camera.
©2024 Bloomberg L.P.