Author | Wan Yi Edit | Fan Zhihui
At the beginning of 2023, the phenomenal popularity of ChatGPT has triggered an "arms race" between global Internet giants on AIGC, and Google, Microsoft, Baidu, Meta, etc. have all ended. The advent of ChatGPT, which can have a natural conversation with people and answer questions, has also turned the abstract and high-end concept of large-scale language models into a "personal assistant" that everyone can use.
Recently, ChatGPT development company OpenAI released the next generation of large-scale language model GPT-4, which is not only the latest AI large-scale language model for supporting ChatGPT and the new Bing, but also supports the new Microsoft 365 feature Copilot, helping users work more efficiently. It is reported that ChatGPT's performance in many professional tests exceeded the "human level", and as soon as the news came out, it once again caused a high degree of discussion about AIGC.
The so-called AIGC (AI-Generated Content) is the content generated by artificial intelligence. In short, artificial intelligence can use existing material libraries to generate content according to user needs. At present, AIGC has been applied to music, painting, audio and other fields, allowing the public to truly perceive the creativity behind the technological revolution.
In the field of music, in addition to supporting music creation, AIGC technology is also beginning to be used to innovate user experience. For example, in foreign countries, Spotify recently launched a DJ mode to allow AI to improve the experience of recommending songs; In China, music platforms such as QQ Music have made many attempts to use AIGC technology. This article will take the AI function application of QQ Music as an example to inventory the imagination space of the combination of music and AIGC.
AIGC, how to change the music experience?
Nowadays, when you open QQ Music, you will find that AIGC has been all over the details of QQ Music, and has made many interesting and innovative attempts from multiple dimensions such as listening experience, visual presentation, and social sharing.
When you click on the recommended song and start listening to it, the AIGC vinyl player with a beautiful design comes into view.
Different from the previous song album cover playback, QQ Music's newly launched AI player is the first visual attempt in the field of AIGC in the domestic music industry, and its principle is to enter keywords in AI tools and let AI algorithms combine various elements to generate a creative player style. Now, QQ Music has launched 6 models such as mechanical armor, snow mountain white, building block games, industrial gray, etc. for users to choose, and it is reported that it will also be combined with scenes, and there are more styles to look forward to.
If the lyrics played are in line with the current mood and want to share them on social media such as Moments and Weibo, the "AI lyrics poster" function comes in handy. Whether it is antique, pop or rap, based on the two models of Stable diffusion and Disco diffusion, in just a few seconds, you can generate a poster corresponding to the style according to the lyrics with one click, saving users the time of looking for pictures.
Behind this is the support brought by MUSE (Music Envision), the industry's first AI music visual generation technology pioneered by Tencent Music Lyra Lab. As the first platform in the domestic music industry to lay out the AIGC field, QQ Music has realized the first large-scale music poster drawing technology in the music industry with its in-depth understanding of audio and lyrics and forward-looking insights into user needs.
It is reported that this technology is also used to generate suitable song covers for a large number of free singles without album ownership in the music library, which greatly improves the user's visual experience, and musicians can also independently make album pictures based on this technology.
With the support of MuseAI algorithm, QQ Music has also developed a playable "AI Dimension Exclusive BGM" function. Users can generate anime-style corresponding images as long as they upload photos of people, and they are also equipped with exclusive BGM.
In addition, in conjunction with the AI technology of Lyra Lab, we can also edit the blessing and select the song by ourselves through the AI beautiful greeting card function, obtain the AI-generated blessing words hidden head lyrics, and sing with the tune of the selected song, and generate customized blessings with one click. I have to say that whether it is a New Year's holiday or a friend's birthday, my mother no longer worries that I will be bald for the thought of blessings, and it is heartfelt and interesting.
When we open QQ Music check-in, in addition to receiving a recommended song every day according to user preferences and listening records, we can swipe the card to view the day's fortune and receive a "Today's Fortune Picture" generated by AIGC. It is reported that the painting is based on the content of the recommended songs of the day and incorporated today's lucky color elements, which is good and good-looking.
For music lovers, the "smart music score" function developed by QQ Music based on AIGC is also very practical and popular.
Since most of the scores on the Internet are incomplete and the cost of screening is extremely high, if you want to learn to play and sing the instrument of your favorite song, the first step is time-consuming notation. Now, thanks to the smart score function, even new songs such as "I Remember" can directly find the score, and the mainstream music scores such as guitar tabs, piano tablatures, and ukuleles are available.
On this basis, QQ Music also makes the still picture music score move, and is the first in the industry to launch the "music score OCR" function, which automatically identifies 10 types of music information such as chords, pitches, and rests in the score based on image recognition methods, and then combines QQ Music's high-precision lyric information to generate the corresponding intelligent music score with one click, helping enthusiasts to play and sing easily, no longer need to stop and manually turn the score halfway.
Different from the general artificial music score, QQ Music's intelligent music score also has a series of functional options such as AB repeat, original sound, metronome, common rhythm selection and editing. Not only that, but we can also share in the QQ Music singing group after learning and get to know like-minded people. This feature was well received and very useful when it was launched.
All in all, whether QQ Music actively combines AIGC to make the user experience more personalized, or improves the learning efficiency of music lovers, it can be seen that AIGC technology has a very high degree of compatibility in the music field, and it also expands the industry's imagination of AIGC and music.
After grafting music, how to open AIGC's imagination?
Previously, the well-known French DJ David Duetta wrote a song about Future Rave in the Eminem style through the AIGC tool during the performance, and recorded it with Eminem's voice, and the audience responded very "crazy", making David Duetta shout "The future of music lies in AI".
And David Duetta is by no means the only one who believes in this.
In the field of music, AIGC has a lot to do in lyrics, composition, and vocal synthesis, just using AIGC to produce functional music can create huge business opportunities, let alone become the mainstream technology to assist music creation, or create imaginative space such as virtual artists. It can be said that AIGC should be the most disruptive technology for the music industry after streaming.
In the face of this historical opportunity, Spotify, three major records and other music companies have actively entered the AIGC field to try to seize the opportunity. For example, Warner invested in AI music creation platform Lifescore, Universal Music invested in AI music creation platform Soundful, and Sony Music independently developed AI-assisted music creation application Flow Machines. There is no doubt that music companies scrambling to explore AIGC has become a new trend.
In China, QQ Music took the lead in gaining insight into the huge potential of artificial intelligence and AIGC and laid it out. In 2019, QQ Music's listening team won the Mirex Audio Fingerprint Competition and broke three world records; In 2020, QQ Music's original Predictive Model (PDM) technology also broke the "Patterns for Prediction" world record, helping to tap potential and good songs, and becoming a standard enabling tool in the industry; It also successfully introduced DNN-based cover recognition technology into the scene of listening to songs, creating a revolutionary first-generation listening and reading system, and the recognition rate increased by 12%.
In 2021, on the basis of QQ Music Multimedia R&D Center, TME established the first audio and video technology R&D center - Tianqin Lab, focusing on audio and video related AI research, and developed the star products of listening to song recognition, intelligent voice editing, audio super-resolution, best mastering, cover recognition, singing scoring and the industry's first MUSE AI music visual generation technology. Among them, the "lyric generation method" and humming recognition developed by Tianqin Laboratory also won the National Patent Award and the Shenzhen Patent Award respectively.
Up to now, Tianqin Laboratory has published more than 10 international top conference papers, obtained more than 500 invention patents, led the formulation of a number of music industry standards, and become a top audio and video research laboratory in the industry. Recently, Tianqin Lab also released three sets of open source datasets for the first time at home and abroad, involving fragment cover recognition, hum recognition and singing evaluation, in order to consolidate the foundation of industrial development and inject new vitality into the upgrading of the digital music industry.
In addition, the virtual human "Xiaoqin", which can be sung and danced by entering the lyrics, is also from Lyra Lab. Based on the support of Music XR Maker System and MÜX Design Center of Lyra Lab, Xiaoqin reflects the comprehensive achievements of Lyra Lab in anthropomorphic dance generation, singing performance generation, music light show and other aspects.
It is not difficult to see that from the underlying technological innovation to the integration of product function upgrades, QQ Music's layout of AIGC has achieved initial results, continuously satisfying and enriching users' audio entertainment experience, and establishing an industry-first-mover advantage.
epilogue
If you observe the functional design changes of QQ Music, you will find that QQ Music not only relies on excellent technical foundation, but also relies on the core concept of continuously improving user experience, and is good at using cutting-edge technology in details to solve the pain points of the B side and C side, so as to continuously lead the iterative innovation of audio entertainment experience.
As user experience expert Sean Gerety puts it, "It's not the technology you use that wows everyone, it's the experience you create with it."
In the face of the ever-changing technological revolution, only by actively embracing new technologies, using them for our own use, and promoting industry innovation can we gain a firm foothold in the new wave. QQ Music's continuous layout in the field of artificial intelligence will not only bring users a more forward-looking and personalized audio entertainment experience, open up the imagination space of AIGC and the music field, but also promote the evolution of music entertainment production ecology.