Plot twist: Warner, which resists AI music, signs a virtual singer

From text, pictures, videos to voice and music, AIGC has penetrated into human audio-visual experience. Even giants that once strongly resisted AI music due to copyright issues had to change their attitude to adapt to the products of the times.

In September this year, Warner Music signed the first virtual singer Noonoouri; in August, Universal Music and YouTube jointly launched the “Music AI Incubator” to solve the music copyright problem after the emergence of AIGC.

Compared with overseas, China seems to be more accepting of AI music. NetEase Cloud Music cooperated with Xiaobing to launch AI music-assisted creation tools and established the first virtual singer label “WOWAIDO!”.

AI-simulated rhythms and vocals are introduced to the public using virtual singers as a carrier. It has changed from the mechanical electronic synthesis sound emitted by the virtual image in the past, and has a voice that is close to a real person. At the same time, virtual singers with changing images and continuous iterations have transformed into AI singers, gradually breaking the distinction between virtuality and reality, and extending more application scenarios and commercial value.

Warner Music signs virtual singer Noonoouri

In September this year, Noonoouri, a virtual internet celebrity with more than 400,000 followers on social media Instagram, signed a record contract with Warner Music and released her first single “Dominoes”. Noonoouri thus became the first virtual singer to sign a major record contract.

The Noonoouri avatar was born in 2018. It is an anime image created by artistic creative Joerg Zuber. The character is positioned as a model. In the past, anime characters served stories and plots to attract fans, but Noonoouri was born to endorse high-end fashion brands. She is active in major fashion magazines and has collaborated with international brands such as Dior, Miu Miu, and Calvin Klein.

Two years after his debut, Noonoouri gained 360,000 fans on Instagram, but in the following two years it became tepid, and the number of fans grew to 400,000 and then stopped. It wasn’t until she entered Warner Music’s artist lineup that Noonoouri unlocked the identity of a “virtual singer”. On Instagram, the MV of the single “Dominoes” has been played more than 250,000 times, making it her most influential work in recent years.

《Dominoes》MV

To make virtual characters move, there are animation, CG and even motion capture technology, but where did Noonoouri’s voice come from? This is about to receive the east wind blowing from AI to this virtual model. It is reported that Noonoouri’s singing part in “Dominoes” was synthesized based on the voice of a real singer with the help of artificial intelligence technology. Both the songwriters and musicians of the work will receive corresponding royalties and distribution shares.

Warner Music, which signed Noonoouri, boycotted AI music in April this year due to copyright issues, but a few months later, the record giant changed its attitude.

Universal Music, another record company that once rejected AI, joined hands with YouTube in August this year to launch the “Music AI Incubator”. It is reported that it will “collaborate with artists and musicians to explore the application of artificial intelligence in music and protect the copyright of music works.” . In addition, UMG and Google, the parent company of YouTube, are also discussing licensing artists’ voices and melodies to train AI models to test events.

It is not difficult to understand the transformation of record companies towards AI technology. Since 2023, AIGC, or generative artificial intelligence, is becoming a part of daily life. Production from graphics and text to video and audio is being transformed by AI and infiltrating into the public’s audio-visual experience. YouTube CEO Neal Mohan once said that in 2023 alone, the number of video views related to generative AI tools on YouTube will exceed 1.7 billion times.

Facing the public’s acceptance of AIGC, record companies have to change their thinking and think about how their accumulated copyright resources can be combined with AI. “Training models - manufacturing tools - stimulating creation” is a good and relatively mature route. In fact, the combination of music and new tools is nothing new. The early electronic music synthesizer is an example.

From synthesizer to AI sound-making platform

In the currently popular “Summer of the Band”, the band “Supermarket” has re-enchanted young music fans of this era with their unique electronic music, and they were the first band to bring electronic music to China.

The electronic music that is gradually accepted by the public is almost all produced by synthesizers. In the early days of electronic music, the synthesizer itself was a piece of hardware, and the sound it produced was expressed through speakers or headphones after being passed through an electrical signal instrument amplifier. Entering the computer age, more and more synthesizer software is appearing. From the perspective of modern pop music production, as long as you own a computer, you are fully equipped to use synthesizer software to produce a complete musical work.

In 2004, musical instrument manufacturer YAMAHA launched the electronic singing voice synthesis software VOCALOID, making electronic music creation more accessible. Not only music, but also human singing can be realized. The creator only needs to input the tones and lyrics to synthesize a human-like singing voice. VOCALOID is what is known as the “V Family” in the domestic two-dimensional circle.

You may not be familiar with the VOCALOID software, but you must have heard of virtual idols such as Hatsune Miku, Luo Tianyi, and Stardust, and their voices come from VOCALOID. Do you still remember the effect of the traditional Finnish folk song “Eva Polka” when it was covered by Hatsune Miku? This cover song became a hit under the name “Onion Throwing Song”.

Hatsune Miku’s cover of “The Onion Swallowing Song” became popular on the Internet

YAMAHA used artificial intelligence, or the basis of AI - big data, when developing VOCALOID. They collected a large number of human voice samples and used neural network synthesis technology to create a sound source library. Users can produce the effect of human singing by setting parameters and calling the voice in the library.

The ability of VOCALOID is used by many virtual idol companies, but the shortcomings of these virtual singers are also very obvious. Since their voices are electronically synthesized, as soon as they open their voices, they have a distinct electronic and mechanical feel. Therefore, the application scope of V family is mostly limited to the second dimension.

But since 2023, virtual people have become more and more real with the iteration of AI visual technology. Adding electronic sounds to them will destroy the realism that was finally upgraded. AI can improve visual effects, and hearing is no problem. Human-like voices trained with large sound models begin to be emitted by virtual humans.

In June this year, Xiaoice, which launched the “Hanzo Forest Clone”, cooperated with NetEase Cloud Music to launch the AI music creation software “NetEase Cloud Music·X Studio” and established a label including 12 virtual singers. WOWAIDO!" These virtual singers can sing both rock and pop, and their timbres cover sweet, cute, melodious and other types of sounds. In the comment area of these virtual singers, someone said, “I can’t believe this is sung by AI.”

It is worth noting that the voices of these AI singers are not simply “copies” of a natural person’s voice. Behind each voice are often synthesized from many different sound sources. It is reported that the sound source is large and authorized.

Solving the copyright issue, AIGC injects “voice” capabilities into virtual humans. The emergence of AI creation tools such as NetEase Cloud Music·X Studio allows musicians to obtain a dry singing voice in a few seconds, and through parameter adjustment, the effect of a song can be presented by the singer.

Looking back on when Hatsune Miku was first born, the purpose was also to solve the problem of producers or arrangers not being able to find singers. Such a “sound container” with an out-of-the-circle appearance, no style restrictions, and subverting the traditional human voice has also inspired producers to keep trying, and this is how the virtual idol industry has emerged.

As “Father of Hatsune Miku” Hiroyuki Ito said: “The more digital works are used, the higher their value becomes.” What creators are doing is “injecting soul into inanimate things.”