A Canadian startup has developed Lyrebird, an audio mimicry program. It could be used in the video game industry
In the age of proliferating fake news, it's hard to really believe what you read on the Net. However, if you think that there is no end to the worst, then you are on the right track. A Montreal-based artificial intelligence startup has, in fact, created Lyrebird. An artificial intelligence capable of imitating anyone's voice.
In the age of social media and the scourge of fake news, the last thing we needed was an artificial intelligence capable of listening to a short audio note from a person and then reproducing the vocal tone by saying whatever the developers decide. In tests, Canadian developers faithfully reproduced the voices of Hilary Clinton, Barack Obama and President Donald Trump. The startup's idea is to offer an API in the future that can allow anyone to use audio mimicry technology. The step between funny and dangerous in this case, however, is very short.
How Lyrebird works
Understanding about the danger of audio mimicry are the developers themselves, who on their site released this statement: "Voice recordings are currently considered as strong evidence in our society and, in particular, in the jurisdictions of many countries.Our technology allows for easy manipulation of audio recordings. This could have dangerous consequences such as: misleading political statements, fraud and more generally any kind of problem caused by stealing someone's identity. By releasing our technology to the public and making it accessible to anyone, we want to ensure that there will be no such risks. We hope that everyone will soon be aware that such technology exists and that reproducing someone else's voice is possible." One of the experts behind the project, Alexandre de Brébisson, says, "The purpose of our research is precisely to show that nowadays even voice can be replicated, so we never have to rely even on what we hear recorded."
The API release
It is not yet clear when Lyrebird will become available to third-party companies. The developers haven't even clarified whether it will be a free tool or not. According to several sources close to the developers though, the service will be launched as freemium. That is, it will have free parts and additional paid features.
Lyrebird will not only copy a person's voice, but it can also be used to modify its own or to control some emotions in the tone of voice, such as anger or stress. These features make the service ideal for video games, audiobook readers, and even text-to-speech tools designed for people with disabilities.