• Innovation through disruptive and scalable technology .
  • Cutting-edge AI research .
  • Accelerating innovations in research and service .
  • We strive for (and achieve) excellence! .
  • “SotA” (State-of-the-Art) .
  • Visual demo of research and service innovation
  • Human. Machine. Experience Together .

Music AI

Producing emotionally touching music service via AI

Background

SK Telecom has a variety of music service platforms and is actively involved in the related business. Recently, a music streaming service, FLO has been launched by taking the third place in the market through various marketing (19.3.). However, the music recommendation services offered by other competitors are only based on simple meta data rather than providing distinctive services. For example, they implement a method of recommending similar music by analyzing previous data on users’ preferences in terms of genre, beat, artist, composer, lyricist, release date, and affiliate company, or a method of removing unselected songs through a Negative Sample Filtering from a new mix tape. In case of AI speaker with an exponential growth in the number of users, the most commonly used feature is to recommend top-1 song by implementing just simple questioning as well as collaborative filtering. Therefore, our next goal is to provide more elaborate and personalized music streaming service and especially, we expect to progressively produce a key technology as 5G service utilizing the AI speaker "Nemo" released in April.

Project

In response to this trend, T-Brain delves on "Singing Voice Translation & Instrument Performance Modification followed by Music Continuation". Especially, we develop AI-based vocal conversion and performance style transfer that can automatically convert existing songs’ components to new versions reflecting desired singer's voice or a specific instrument performance style based on the user's preference. When these technologies are commercialized, it is possible to listen to the popular songs with favorite singers’ voices and automatically convert the karaoke accompaniment to a sound file or a music score by changing the style of accompaniment to various keys and tempos. In addition, music continuation service, which automatically generates music and accompaniments referring to the main melody motif, will be available. Furthermore, we also plan to implement “5G Realistic AR/VR Game Ambient Sound Generator” and “Visual Navigational Recommendation System.”

Conclusion

With our research on various music AI technologies, we aim to develop a service providing not only rooms for the enjoyment but also an emotional touch to service users. Singing Voice Translation will be the key technology to differentiate music streaming service among many other competitors; Performance Style Transfer will lead to new services, such as automatic accompaniment production, tempo conversion, and music continuation. In addition, we would like to propose a new 5G service which can provide high-quality sounds for both realistic and real-time game service. Through visual navigational recommendation system, we will provide more personalized, rich and user-friendly environment for contents recommendation to music recommendation service users.


  • Bob
  • Ray
  • Daniel