When OpenAI demonstrated ChatGPT's natural conversational abilities a few months ago, even the most rigorous AI critics could not help but be impressed. Watching the presenter converse with the AI as if it were a friend sitting next to him seemed like something from the future.

And it turned out to be from the future. Because the feature was not ready for launch due to safety concerns and other issues. But OpenAI is finally ready to let the public try out its Advanced Voice feature. The company has announced plans to make an alpha version of ChatGPT Advanced Voice available to a small subset of ChatGPT users.

In an X post today (July 30), the company said, "We are beginning to roll out Advanced Voice mode to a small group of ChatGPT Plus users. "OpenAI did not say how many users will have access to the feature, but some X users posted screenshots showing that it is available. If you are using the free version of ChatGPT, don't expect to have full-fledged conversations with your AI companion anytime soon.

As for safety issues such as the tool being used to mimic real people's voices, OpenAI says, "To protect people's privacy, we trained our models to speak only in four preset voices and built a system to block output that differs from those voices and We have also built a system that blocks output that differs from those voices."

If it is not sure whether it has received access, OpenAI says." This alpha user will receive an email with instructions and a message on the mobile app." If they have not received that email, they will have to wait for another wave.

This is just the first wave of gaining access, OpenAI added in its post, "We will continue to add more as we go along and will have access for everyone who signs up for Plus in the fall."

One feature that will not be ready immediately is video and screen sharing.

Ending the barrage of X posts, OpenAI spoke about the future: "What we learn from this alpha version will help make the Advanced Voice experience safer and more enjoyable... A detailed report on GPT-4o's capabilities, limitations, and safety assessment will be shared in early August. We plan to.