Microsoft made an AI voice so real, it’s too dangerous to release
Microsoft has a new AI toy — and no, you can’t use it.
Researchers at the company revealed in a research paper (spotted by Live Science) that they’ve built an AI voice synthesizer that can create believable, realistic human voices straight from text prompts.
Called VALL-E 2, the research paper claims it’s the first time “human parity” has been achieved in a text-to-speech system.
Of course, there’s no way for any of us to know that for sure because regular people can’t use it. Microsoft’s blog post is very clear about this.
“This page is for research demonstration purposes only,” Microsoft’s website said. “Currently, we have no plans to incorporate VALL-E 2 into a product or expand access to the public.”
Unfortunately, there aren’t audio samples available, so you’ll have to trust the researchers on this. However, the blog post does offer numerous charts and technical terms to explore. If it’s as good as they say, though, maybe it’s for the best that we can’t hear it.