September 4th 2017
After four months of hard work, we are happy to announce that we are launching a beta version of Lyrebird.
This beta version lets anyone create a digital copy of their voice with only one minute of audio. Beta testers who record themselves can then use their digital voice to generate any sentence they like. Our tech is still at its early stage and these artificial voices are still basic and imperfect but they capture certain characteristics of a voice identity. We also want to stress that one minute of audio is the minimum amount that we require and more audio recordings will definitively improve the quality of the digital voice. We hope to make rapid progress in the coming months and we will regularly update user voices with our latest improvements.
This technology will create a wide range of new applications. We’ve received interest from hundreds of developers targeting many different use cases. We are going to start a private beta with 10 selected developers, which have the following profiles:
When we first launched four months ago, many of you were very excited and curious about the new applications that this technology would make possible but some of you also raised a few legitimate concerns about fraud and fake news. As pioneers of the technology, we take those questions very seriously and we describe our ethical stance in this statement. The main takeaway is that your digital voice is yours and no one else can generate from it unless you give them your consent. With the exception of the voices of Obama and Trump that we use to raise public awareness (see our video and our demo).
Lyrebird is building cutting edge speech synthesis technologies. We’re backed by some of the best investors in the world: A16Z, Redpoint, YCombinator, SVangel. We need great people to help us build this technology and guide its development. If you’re interested, send us an email at email@example.com.