The Voicery initiative, powered by Y-Combinator, uses artificial intelligence to generate artificial sound from the brand.

Although the San Francisco-based Voicery initiative is only a few months old, the CEO and co-founder of the initiative, Booby Ullman, said they have applied to hundreds of companies that want to create their own voice. Because Voicery offers something that many companies do not know is what they need until five years ago: a personalized digital voice that sounds like a real person, not a computer.

Unlike the dull voices we hear in today's customer service calls, Voicery's artificial intelligence sounds make us feel like we're talking to a real person. While the interaction we have established with companies is shifting towards the vernacular from day to day, the tone, quality and rhythm of the voice used by a company become the 'new face' of that brand.

Sound can be an extremely powerful brand vehicle. For example, you have noticed how strange Alexa is in the ears when joking. Because it is extremely difficult for artificial sounds that imitate human speech to give a convincing feeling with that robotic voice and pause. Most of these computer generated voices use speech synthesis method called concatenative model. In this method, a voice-over artist is spoken about 200 hours of talk, digitally subdivided into small pieces of voice and rearranged to tell what to say if necessary. Voicery's model works differently. The system, which only listens to a few hours of talking by a voice-over artist, then trains a deep nerve network to mimic that person's voice. The entire process lasts two weeks. It takes four days to construct a single artificial voice neural network model. At this point, Voicery has three ready-made artificial voices.

By entering Voicery's web site you can solve the test which requires you to guess which voice belongs to the machine and which voice belongs to the machine. You will notice that it is quite difficult to distinguish between machine and human voice when performing the test.

Voicery's technology can be extremely useful for companies that want to make sure their brand is consistent across all interfaces. For example, would not it be nice if the monotone voice that stood on the other end of the line when you called your insurance company responded like a real person? Even if you speak with the voice of an artist you like?

The credibility and charisma of voice can be more important than you might think for companies that want to relate to their users. For example, a health care practice might be more likely to feel a stronger affinity for that brand if you speak with the voice of a real-life friend rather than a robotic "You can!"

Whose voice can be imitated?

There are also ethical debates about what types of voices can be articulated by entrepreneurs, as well as by artificial intelligence that produces fake videos. Ullman wants to draw a clear line between the types of voices that Voicery will and will not produce: "As such tools get better, ethics need to be more cautious." Voicery's website states that the voice will never be imitated without someone's permission.

Voicery will also serve companies that want to produce their own voices. This B2B service will be Voicery's core business model. As computers move out of the screens and into the areas we live in, we want to communicate with them in a way that does not make them feel robotic or do not come to the ear. At this point, Voicery's technology seems to be working quite a bit.