A new era in music begins with these data-driven virtual entertainers.
Though it may come as a surprise to some, the concept behind “virtual entertainers” like pop star Hatsune Miku or Instagram influencer Lil Miquela has been around for decades. Fictional musical groups like Alvin and the Chipmunks and Jossie and the Pussycats, though vastly different from the types of acts we see today, lead the way for future fictional musicians, including The Gorillaz, Dethklok, and numerous others.
Despite the outrageousness of some of these characters, the fact remains that behind all of these performers are teams of human musicians responsible for creating their signature sound. This is where the Auxuman’s come in.
Created by virtual entertainer development studio Auxuman with funding provided by Betaworks, the Auxuman’s are a collective of AI-driven virtual musicians. Similar to Hatsune Miku and Lil Miquela, this supergroup of 3D animated performers each feature their own unique sound, style, and appearance; but whereas Miku and Miquela’s music is produced by a team of professional engineers and vocalists, the Auxuman crew use a combination of internet data, artificial intelligence, and generative tools to develop their own original compositions.
Last month the collective dropped their first full album online, featuring music from each of the five artists: Yona, Mony, Gemini, Hexe, and Zoya. Auxuman has instilled each character with their own personality and emotions based on articles, music, and other influential sources. The idea, according to the developer, is to create AI personalities inspired by human behavior; how would an artificial being behave while imitating a human being?
Yona, one of the first characters developed and leader of the group, was “fed” the works of Margaret Atwood, as well as publications pertaining to teenagers and city life. In the end, Yona developed an innocent, romantic personality unique to that of her other “bandmates”.
“We started with one character with one personality named Yona. She’s usually the lead of this story. And what we found out was that the sort of pipeline that we created could be shaped in many, many different ways in terms of the psychology and the personality behind these characters and what they want to say”, said Ash Koosha, Founder and CEO of Auxuman, while speaking to VRScout.
“So using one character, I think it wouldn’t necessarily diversify the outcome because the nature of storytelling is that you want to become familiar with one set of personality traits and psychological behavioral patterns. And we expanded it to more tests and we started saying, “Oh, ok, let’s not choose this type of personality. Let’s go for more of an aggressive, angsty type of character”. And then Hexe was born. Then, more and more, we discovered more interesting areas in terms of personality. Each one of them have a case file for their personalities and that drives what rules they follow”.
While the “auxiliary humans” are responsible for generating a majority of the sound, Ash and his team are still on-hand to help each personality finalize their work, sort of like a producer for a real-life band.
“What we’ve done is that we have a series of engines altogether. Each part is responsible for generating part of the music. So let’s say, what does a song need? The lyrics, melodies, rhythm, sequences, audio texture; all of this generated by machines. We act as curators and engineers that put all of these pieces together. So it’s fully digitally generated.”
When asked about future plans for their lineup of artificial artists, Ash elaborated on the team’s interest in game streaming as well as eventual live performances.
“I think there is a lot of promise in game streaming and cloud gaming where we can actually merge the lives of these characters into platforms that people can access. So it’s gonna be a bi-directional scenario. I think their presence is going to be very transmedia.”
For now, you can check out the group’s first 10-track album, Auxuman Vol. 1 for free. Songs include “One” by Yona, “Marlene” by Mony, “IF Sad” by Gemini, “Never Alone” by Zoya, “Subject” by Hexe, as well as five other AI-generated compositions. While a majority of the “singing” could be considered more spoken-word poetry, the overall complexity of the instrumentals and lyrics is extremely impressive for an AI-generated personality.
Each virtual being has its own recognizable sound, adding a surprising amount of diversity to the album. Not a bad debut for the digital collective; it’ll be interesting to see what they bring to the table with their sophomore follow-up.
Feature Image Credit: Auxuman