Speech recognition and generating sound

What is it?

AI can turn speech into text and text into speech, recognise sounds and create music or a voice that sounds like a real person. Quality is now so high that a synthetic voice is often impossible to tell from a real one.

An everyday example

Your phone transcribes a voice message, and your car recognises the command “call home”.

Why does it matter?

Speech technology makes things more accessible for people with impaired sight or hearing, but it also enables scams with faked voices.

A common misconception

That a familiar voice proves who is speaking. If there is an unusual request, such as for money, check the caller another way.

In more detail

What is speech recognition?

Automatically turning speech into text — often with mistakes when there is background noise, a dialect or unusual names.

What is speech synthesis?

Creating spoken audio from text or other instructions.

What is voice cloning?

Creating a voice that sounds like a particular speaker. This raises questions of consent and abuse.

Is an audio recording proof?

Not on its own. A voice can be generated or edited; you have to check where the recording came from.

Remember: A voice that sounds familiar is not proof of who is speaking.