AI singer creation software online, how can musicians co-create with AI?

On July 6, the first virtual singer label WOWAIDO! is about to be founded. This label composed of 12 AI singers has released 3 mini albums after its debut. In the NetEase Cloud Music comment area, based on the high concurrency of artificial intelligence, they interact with fans "one-on-one" and even show different personalities.

AI singer Ruoxi (codename F970) performed brilliantly, because his voice resembles Faye Wong, and a song "Ruoxi You" produced by a Vietnamese team has gained 999+ loves in just a few days. At the same time, music creators use XStudio (AI singer music creation software) to create a lot of her voice.

A new era of AI seems to be beginning, but it has also caused public unease. How are the rights and interests of live artists protected? Are humans going to lose their jobs in the field of singing?

AI singers debuted in a group

As AIGC technology (generative artificial intelligence) pushes the industry into a dynamic new generation, all walks of life are constantly gaining inspiration and momentum in the wave of AI innovation. On the 15th of last month, Xiaoice Company and NetEase Cloud Music announced a strategic cooperation to launch the AI singer music creation software "NetEase Cloud Music · XStudio", free for musicians and music lovers. The software has the above 12 AI singers built-in, musicians input songs, within 3 seconds can get professional singer-level AI singing dry voice, and can also adjust parameters to achieve AI singer's glide, transliteration, vibrato, biting words, etc., delicate interpretation of the work effect. This is also the first product of its kind jointly created by a global mainstream music platform and an artificial intelligence company.

The person in charge of the product of the Xiaoice company told reporters that the current sound library of the software can basically cover most of the music creation needs, and will continue to expand the AI singer camp to include rap and other styles.

The list of existing AI singers includes Shoujiao Xiaoice, He Chang, Chen Shuiruo, Xu Mengtian, Ruoxi, Cui Xuan, Yan Zeyu, Xiaoyu Tofu, Xia Yubing, Little Cousin, Chen Ziyu, Lu Lu. Among them, there are many "AI singer celebrities" with equivalent works. For example, He Chang (code name F11), who has cooperated with Ma Boqian, Bian Ziyan and Xiaoke and other well-known musicians, launched the promotional song "Waiting for You in Chengdu" for the Chengdu Universiade and "Only You Have No Other" for the Beijing Winter Olympics. The girl Xiaoice (code name F002) is the "head of the family" of the Xiaoice company, has published a collection of poetry, and holds many college diplomas such as the Central Academy of Fine Arts and the Shanghai Conservatory of Music.

It is reported that song production usually includes several stages such as lyrics, composition, arrangement, and recording. In the past song production process, live artists generally did not intervene until the recording stage. And AI singers can help in the stage of lyrics and composition.

Sun Yujing, founder and artistic director of the fantasy animation music team, said that songwriters may not have the vocal conditions to be good at singing. The advantage of AI singers is that they are "on standby" in real time and have a constant state, presenting the singing results of the lyrics to the creator, helping the creator judge whether it meets the expectations and adjust it in time, thereby increasing the upper limit of creation.

How does this AI music software work? How to apply rich cutting-edge artificial intelligence Xiaoice black technology to music production? The person in charge of the product of the Xiaoice company told reporters that the software is equipped with a number of technologies. Xiaoice singing model and consistent super natural voice technology, AI singers can achieve singing effects that are indistinguishable from real people. Streaming rendering singing synthesis technology allows singing voice generation to be completed in less than 3 seconds. In addition, "NetEase Cloud Music · XStudio's AI tracks support synthesis up to 30 tracks, which means that every musician has a "30-piece" choir.

Who protects the rights of live artists?

Not long ago, "AI Sun Yanzi" was born on various music platforms. An AI speech conversion project developed by Chinese singing synthesis, creating "She". "AI Sun Yanzi" has more and more song lists and podcasts, bringing ethical discussions. Music creators have the question – do third parties have the right to copy the artist's own voice without permission?

Especially this time in the "NetEase Cloud Music · XStudio" appeared in Ruoxi, and some netizens used her voice to make a song "Red Bean" and upload it to station B, which resembled Faye Wong's voice to amaze everyone. Are Ruoxi and other AI singers suspected of infringement, and how are they different from AI Sun Yanzi? At the legal level, who will protect the rights and interests of live artists, and how will the problem of "Sun Yanzi" be solved?

In response to this phenomenon, Xiaoice company said that the problem with "AI Sun Yanzi" is that a third party has privately copied her voice without the permission of Sun Yanzi herself, so that Sun Yanzi cannot control the use of her AI voice and cannot obtain relevant benefits. Many AI singers in the market now have similar problems.

Xiaoice company's product leader said that XStudio put AI singers under the sun for the first time, and hoped that this move would provide a benign development reference for the future of AI creation. The software contains high security and privacy protection policies, all AI singers have undergone rigorous data training, and all creations can be traced.

For Ruoxi, whose voice resembles Faye Wong, the above-mentioned person in charge further explained that there are generally two types of AI singers, which can be distinguished from sound sources. One is the "cloning" of the voice of a single natural person; One class is synthesized by the training of multiple natural human sound sources. The sound sources of these AI singers in XStudio have no copyright issues, including Ruoxi. After adjusting through software, netizens found that her voice was similar to Faye Wong, but Ruoxi's sound source was not Faye Wong, and there was no problem of "stealing" privately.

At the same time, Xiaoice company told reporters that live-action singers have taken the initiative to authorize companies to "clone" their voices to retain their peak state. Singers can decide what AI voices are used for and make related benefits. When more and more "genuine" voices appear in the market, good money drives out bad money, which may solve the problem of "Sun Yanzi" and give music creators more choices, which may be the first step in the legalization of "AI Sun Yanzi".

How can musicians collaborate with AI to co-create?

On top of the bottom line determined by the law, the new generation of AI, musicians and AI collaborative creation will bring more possibilities. "NetEase Cloud Music· XStudio" released on the day of its release, musician Song Chengrong posted a circle of friends, and he mixed for AI singers last year. Song Chengrong said, "In the future, the efficiency of AI will become higher and higher, and the quality of music will be better and better." It's a good thing that AI has actually changed our lives and even set new standards. But for people who don't know music and still make music, it must be a bad thing. Those who can't make music and AI are probably embarrassed to sing. ”

"With the current technical standards, I believe that it will not be immediately distinguishable to be an AI singer within four sentences." Musician Sun Yujing told reporters that the emergence of AI music creation software allows creators to feel the final presentation effect of songs to the greatest extent. "In the past, without AI production software, we would have a problem making demos – we had to sing them themselves. The producer is very professional in the theory and production of music, but the singing effect may be worse. ”

For example, Sun Yujing once created a rock song, and he had to audition at the beginning of the creation to find a feeling, but his voice was not suitable for this genre. Choi in XStudio has a rock voice, which allows him to feel at least 80% of the composition before the recording stage, greatly improving the success rate of musical works.

For young teacher Li Zhaoyang, a music producer who likes to create children's music, the creative pain came from not finding the right singer. Li Zhaoyang introduced that high-quality children's songs can not only promote children's language learning and intellectual development, but also make children have aesthetic awareness and subtly affect their values. At present, the number of good children's music on the market is limited, the range of choices for parents is small, and the demand and production capacity are extremely mismatched.

"This kind of music needs to be performed by young children, but there are not many live-action singers." Li Zhaoyang said that after listening carefully to the voices of AI singers on the market, he found that the girly Xiaoice voice in XStudio was the closest to a child's voice he found in similar products. He used this sound to compose and published the work "Rhymes of Kasion". The track was nominated for the 5th Music Awards of the Singing Committee.

But AI is not omnipotent. Sun Yujing admits that the human advantage is more obvious in the recording stage, because the communication between people is more direct, and he can let the live singer quickly adjust the interpretation of a certain detail through personal demonstration. For AI singers, the details can only be changed by adjusting parameters, which tests the creator's familiarity with the software. In addition, there are not many AI singers on the market who can meet the mature standard. Therefore, in his opinion, AI singers and real singers do not threaten each other, they are indispensable in the music creation process.

Text/Reporter Zhang Zhiyi