After previewing its new Audiobox generative AI project last month, which can replicate your voice using just seconds of sample audio, Meta has now launched a publicly accessible demo of the process, which enables you to create custom audio samples based on your own voice and/or text prompts.
As explained by Meta:
“Audiobox is Meta’s new foundation research model for audio generation. It can generate voices and sound effects using a combination of voice inputs and natural language text prompts - making it easy to create custom audio for a wide range of use cases.”
And it’s pretty good, though also pretty freaky.
First off, the Audiobox demo provides access to a range of elements, including voice descriptions, sound effect generation, audio editing, and more.
The main sample use case is custom audio generation based on text prompts
As you can see in this example, Meta has provided a few basic Audiobox test elements, including a text-to-speech process that enables you to generate custom audio based on any text input.
There are only two system voices to use in the demo, “Alice” or “Emily”, but it gives you an idea of what the process can do, in terms of translating custom text into alternative audio streams.
You can also add custom sounds to your sample, again based on text prompts.
But the freaky part is the option to generate your own voice:
It definitely sounded like me, and I definitely did not like it. This, of course, could end up being a significant concern, in terms of potential misuse, and Meta does make you agree to a range of terms and conditions of use before trying it.
But it’s pretty good, pretty accurate. Uncomfortably so.
It’s the latest in Meta’s growing slate of generative AI tools, which are set to see varying use cases in its apps over the next year. And while Meta’s still building in security parameters, it is a concern that it’s making such tools more widely accessible, even with the various terms of service conditions attached.
I’m not sure the world is ready for it, in all honesty, and with various elections coming up, I can imagine a wide range of issues arising from potential misuse.
But AI development is moving fast, and Meta wants to keep up.
If you want to see how good it is at replicating your voice, you can test it out for yourself here. Call your partner and play the snippet down the line, see if they can pick that it’s not actually you.