AI-generated realistic voice for ChatGPT now accessible to plus users

OpenAI has introduced early access to ChatGPT’s Advanced Voice Mode for select ChatGPT Plus members, featuring a highly realistic AI voice.

This technology, built on GPT-4, transitions from the previous three-model system to a single multimodal model, offering greater responsiveness and realism. It delivers a more natural-sounding voice, can detect emotions in speech (including singing), and can handle interruptions, allowing for real-time story changes.

The current version is an alpha release, meaning it is in the early stages and lacks screen sharing or video features mentioned earlier this year.

OpenAI's debut follows a recent controversy involving actress Scarlett Johansson, who claimed that the company’s “Sky” voice preset sounded very similar to her AI character in the movie *Her*. OpenAI denied any intent to imitate Johansson, but the Sky voice is notably absent from the current ChatGPT release. The company also stated that generative AI can no longer impersonate real people and is restricted from creating outputs that deviate from the stock voices. Additionally, the system is prohibited from generating music or other audio that could infringe on copyrights.

The feature might not be widely available for some time, and there's no assurance it will be accessible to free users. This launch coincides with advancements in existing voice assistants. Apple is enhancing Siri with Apple Intelligence, which will be available on the iPhone, iPad, and Mac this fall, offering more comprehensive answers, performing additional tasks, and understanding on-screen context. Google plans to integrate Gemini more deeply into Android and Chrome OS. Despite significant differences, OpenAI may find itself competing with major mobile companies for users' voice commands.