Shulou(Shulou.com)11/24 Report--

In the same AIGC, some people are using AI to generate sketches, others are using AI to generate music, and AIGC is also generating music, some are imitating humans, and some are reproducing classics. AIGC is a tool that generates different values in the hands of different people, which can be trillions of dollars. [1]

In front of AIGC, who else does not kneel down near the end of 2022, AIGC has become one of the most concerned technical keywords this year. With only a few sentences of description, AI can associate and paint, and through the most direct and rapid visual feedback, it will have an impact on human's conventional understanding of art.

Belong to the less good AI painting through the previous article "when did the AI factory workers start to move bricks?" You probably already know what AIGC is and the current investment situation. If you look to the future, as Gartner calls "one of the five impact technologies of 2022", AIGC will create 10 per cent of all data produced by 2025. [2]

Gartner released the 2022 technology maturity curve [3] although the current visual competition of AIGC is more full, faster, and more eye-catching, but music as an important carrier of human culture and emotion, AI is actually transforming the industry-only more imperceptibly.

Tags, algorithms, neural language recognition and other functions, almost become the standard music App, they can automatically generate a playlist and push to the user. Now when you open QQ Music, many people should throw the choice directly to "30 songs a day" and "personality recommendation".

If anyone wants to experience the feeling of being ridiculed for their taste in music, AI can also provide a thoughtful service. At the end of 2020, online magazine "The Pudding" launched an artificial intelligence application "How Bad Is Your Spotify?" [4] millions of Pitchfork comments, reddit posts and record store recommendations were entered into the app's database, and enthusiastic Twitter users even flooded the Pudding site by analyzing users' Spotify playlists and mocking their "rudimentary" taste in music or being "uncomfortably obsessed" with a singer.

AI's transformation of music does not stop at interest analysis or taste appreciation. The 2014 American TV series "Scorpion,S2E8" tells a story that was quite science fiction at the time: a blog musician wrote a program that could constantly create "Divine Comedy," but was killed by another songwriter who used the program to write songs for profit.

Eight years later, thanks to the efforts of the big platforms, the fantasy in Project Scorpio quickly became a reality. AI is finally going to really do something about the music creation process!

They are all playing AIGC, and they all have a bright future. In June 2016, the world's first artificial intelligence to create 90-second piano music was born, which comes from the "Magenta" project jointly launched by Google and the California Institute of Arts.

Since then, there has been more and more AIGC music, and the underlying technology has been divided into two ways.

One way is to "compose music automatically". In 2017, Sony CSL (computer Science Lab) artificial intelligence program FlowMachines released "Dad's car", which is the result of AI composition, claiming to be "the first complete song written entirely by artificial intelligence", while Microsoft's AI released an AI song called "Joyful Time in The City" by analyzing and learning dozens of Christmas carols.

Compared to the 90-second composition, the full AI automatic composition is a big step forward. But they still need to draw inspiration from the vast sea of human historical music, which is an extension of human artistic inspiration. Works like "Dad's car" are composed and arranged by AI, harmony and lyrics are done by human composers, and "City Happy time" also needs human assistance. As a matter of fact, this kind of music is not "created entirely by artificial intelligence". It is more likely to be self-inflamed in a technical circle, and the actual acceptance of the audience has to be called into question.

Another way is to focus on "AI automatic synthesis of music works". For example, the sound synthesis technology engine "Ayane" developed by Tencent Music Apollo Sound Lab highly reproduces the singer's voice characteristics and reproduces the artist's classic voice. The first AI singer was released in July this year. Based on Yang's voice synthesis, it only takes about 10 minutes for the singer to reproduce the original sound, and even with "emotion" in the performance. Since then, mature works include "unsent letters" (reproduction and reproduction of Teresa Teng's voice), "May you be gently waited for by the world" (reproduction and reproduction of Anita Mui's voice).

The difference between "automatic composition" and "Ling Yin engine" is that "automatic composition" imitates music. "Ling Yin engine" uses a self-designed deep neural network model, coupled with the existing sound style of "synthetic music". The latter is not shy about using the best works of art in human history and replicating these artistic styles.

This is another development direction of AIGC music-instead of letting machines create original music, machines can help people create music and do more to change the way we experience music in the past than simply outputting songs.

Douglas Eck, the technical director of Google's AI music project Magenta, said in an interview with Science that AI has become a chess master, but now it has become a helper of human chess skills, as well as music. no matter how fast the machine learning speed is, humans can adapt and use AI technology to assist creation and create new ideas at a faster speed.

To put it bluntly, AI doesn't have to replace humans, and it's good to be a handy tool. What needs AI tools to generate content most may not be in reality, but in the virtual world.

That is, how to completely digitize the "people in the middle".

Today, the music of AIGC, which uses AI to the cutting edge, is different from the automatic speech synthesis technology in the past, but has a higher reproduction scale. The AI sound rendered by the engine is closer to a real person in timbre and tone, and is more expressive and "emotional" in singing. The current "Ayane" technology not only allows AI to sing songs that already exist (officially known as "synthesis"), but also allows users to customize composite songs written, composed and sung by AI with one click.

In this environment, the problem people face is: how to make meta-universe digital man sing and dance Rap to play basketball?

This is not easy, because the video world not only needs music, but also requires the sound, mouth shape, expression, movement and other elements of music (musical instrument / human). In richer scenes, it also requires dance, mirror operation, and lighting.

If you use the "Music XR Maker system" of Tencent Music Tianqin Lab, you can fully meet these requirements, and the application scenarios are diversified. The system's first virtual human violin was launched in October this year, using AI-driven way to generate dance performances and singing elements based on music, with the aim of improving operational efficiency and virtual human authenticity and fun. This also provides a better solution for scenarios such as meetings, customer service, news broadcasts, etc. It is understood that soon there will be a "super singer" who has come a long way from the "next generation" and has been selected layer by layer to make her debut in TME. Maybe she will subvert our imagination of singing and become the sky falling purple star that opened up the "first year of music virtual idol".

"unlike the customized head virtual idol with a large investment, we hope to build Music XR Maker into a platform for producing digital people with high efficiency and low cost, while at the same time enabling digital people to have stronger real-time content production capacity and make it possible for the interactive combination of any music, dance, stage and other elements." Zhou Wenjiang, vice president of technology of QQ Music, said.

Meta-universe's AIGC music also has to make money. In 2022, Zuckerberg burned $10 billion of meta-universe social platform Horizon Worlds (Horizon World) retention touching. On October 16 this year, Meta internal files were exposed, which showed that the number of monthly active users in Horizon World was less than 200000, far from meeting the original 500000 expectations, and the interest lasted only one month.

It is funny to say that as a virtual world, what meta-universe lacks most is people, and the most difficult thing is to create a good soil to attract people to create together.

In comparison, another domestic meta-universe case is more pragmatic. On July 16 this year, TMELAND, a virtual music carnival owned by Tencent Music, created an immersive virtual 3D performance with Pepsi. This virtual music performance provides content and theme for meta-universe experiments, and this commercial partnership also provides a "standard template" for future virtual scene marketing. This may be the first big commercial customer to pay the bill in the domestic AIGC field. Many brands are blessed to the content co-construction of meta-universe, which will also enrich the ecology and form a marketing scene with vitality.

Throughout the music industry, from the early era of 1.0 records, the output of music content was concentrated in the hands of a small number of professionals with resources; in the era of 2.0 streaming media, more music practitioners participated, and the recommendation and distribution mechanism gave them opportunities to show. with the advent of 3.0, AIGC even allows music consumers to have the ability to reverse output and participate in co-creation. Continue to strengthen the investment in AI, the future of the music platform may be a more fertile soil than imagined.

Enough high-sticky users, people's never-tired appreciation of music, formed payment habits, progressive co-creation tools (not directly involved in creation can even avoid copyright disputes), and diversified scene collisions. who says the music platform is not a better soil than a gaming meta-universe?

"meta-universe has never been achieved overnight, we can use our mature technology to achieve a basic version, and then continue to improve various technologies to improve the meta-universe experience." With regard to the more possibility of music + AI, Dong Zhi, head of computer vision at Tianqin Lab, said: "our core competence in AI driver, they all have one thing in common, that is, to generate the basic elements of the virtual world through music and AI."

About AI art, from AI writing to AI painting to AI music and video, AIGC has always faced some controversy: if the generated content is not good enough, then what is the meaning of its existence? If the content it generates is good enough (and this may be inevitable in the future), will it provoke a new round of career replacement panic? Will human spiritual entertainment life be taken over by AI?

This may be a series of persistent topics. Technical optimists often say that the quality of tools depends on the purpose of the user.

In fact, what the Internet has been doing is to make it easier for more people to enjoy the pleasures of technological upgrading, as well as in the field of music. If we open up Ayane engine, Music XR Maker and other tools to all musicians, production organizations or content companies in need in China, and then help listeners enter the music experience of the combination of reality and reality as soon as possible through Music Metasmology and AIGC, it may even become a powerful tool jointly created by the whole people in the music field in the future, so that everyone can make music and enjoy music according to their own imagination, then it may be a more desirable future.

References:

[1] Generative AI: A Creative New World 2022.9.9 https://www.sequoiacap.com/article/generative-ai-a-creative-new-world/

[2] Predicts 2021: Artificial Intelligence and Its Impact on People and Society https://www.gartner.com/en/documents/3995104

[3] What's New in Artificial Intelligence from the 2022 Gartner Hype Cycle https://www.gartner.com/en/articles/what-s-new-in-artificial-intelligence-from-the-2022-gartner-hype-cycle

[4] How Bad Is Your Streaming Music? Https://pudding.cool/2020/12/judge-my-spotify/

[5] Science:How Google is making music with artificial intelligence https://www.science.org/content/article/how-google-making-music-artificial-intelligence doi: 10.1126/science.aan7216

This article comes from the official account of Wechat: fruit Shell hard Technology (ID:guokr233), author: dry Ma Duck, Editor: Yimeng

Welcome to subscribe "Shulou Technology Information " to get latest news, interesting things and hot topics in the IT industry, and controls the hottest and latest Internet news, technology news and IT industry trends.