Improvements in voice mimicking technology could lead to a new form of "vocal terrorism", scientists are warning.
At the current rate of progress, researchers believes that in 10-15 years computers will be able to perfectly mimic human speech after hearing just one sentence.
Aside from the obvious implications for Karaoke and prank phone calls, scientists also believe the technology could have serious repercussions in the fields of fraud and terrorism.
One of the scenarios postulated envisages a phone call from someone who the customer believes to be their bank manager asking for account details, which could then be used by fraudsters to empty their bank account.
Scientists also speculate on a type of terrorism whereby an official's "voice" is broadcast over a country's communications network, carrying false instructions or information to citizens.
Past attempts to replicate human speech on computers has failed because the technology relies on copying sounds, and so misses out the subtle variations in pitch and rhythm that define natural speech.
However, scientists speaking at the British Association Science Festival say they have overcome this limitation by scanning the vocal tract with an MRI machine and copying the alterations in the muscular structure which occur when speaking.
So far the technology has only worked with vowel sounds, but consonants are the next step they say.