A few months ago, we reported about Adobe’s new software VoCo that allows you to take audio recordings of someone’s voice then doctor them, making the speaker say words or phrases that they never actually said.
VoCo was nicknamed “Photoshop of Speech,” for that is basically what it is, a software editing tool that can manipulate digital files for both legitimate creative ends and willful deception.
In fact, VoCo is being labeled as a security threat by some media and cybersecurity experts due to the various ways it can be used to “put words in someone else’s mouth” – an obvious nightmare for politicians, celebrities, lawyers and journalists.
VoCo’s caveat is that it requires at least 20 minutes of audio samples before its algorithm can successfully mimic someone’s voice.
Now, a new start-up wants to one-up that and claims its software can copy and synthesize anyone’s voice with just a minute long recording.
Lyrebird – Copy the voice of anyone
Montreal-based start-up company Lyrebird is set to launch its new service that uses deep learning artificial intelligence and algorithms to mimic anyone’s voice from even a minute worth of audio samples.
The company claims that its software can compress anyone’s voice “DNA” to generate a unique key. This key can then be used to generate and synthesize speech with the corresponding voice “DNA.”
Aside from creating artificial speech, users can even control the emotion of the computer-generated voice to express human subtleties like anger, stress or sympathy.
The developers said that their processing clusters are powerful enough to generate 1,000 real-time sentences in less than a half a second.
To showcase how accurate their voice-mimicking technology is, Lyrebird’s website has a number of audio demos with artificial talk from the likes of Donald Trump, Barack Obama and Hillary Clinton.
Lyrebird stated that since it’s “the first company to offer a technology to reproduce the voice of someone as accurately and with as little recorded audio,” it is aware of the potentially dangerous consequences of this technology such as misleading diplomats, fraud and identity theft.
They also raise the issue of the validity of audio as evidence in court. By releasing their software publicly, they hope that everyone will be aware that such voice copying technology is readily available, which can invalidate the use of audio recordings as evidence in the future.
“Voice recordings are currently considered as strong pieces of evidence in our societies and in particular in jurisdictions of many countries,” Lyrebird wrote on its Ethics page. “Our technology questions the validity of such evidence as it allows [you] to easily manipulate audio recordings.”
Creepy usage aside, Lyrebird stressed that its technology has legitimate practical uses as well.
The company is pushing its software as suitable for companies looking for new speech synthesis solutions. Their technology can then be used in a wide range of applications such as personal assistants, audio book reading with celebrity voices, speech synthesis for the disabled and for voice-overs for movies and video games.
What do you think? Are these voice copying software and services cool technologies or are they dangerous security risks? Drop us a comment!