KKCompany | #KKnowledge 生成式音樂嶄露頭角，人人都能成為音樂創作者？
最近歌曲生成在音樂創作社群中嶄露頭角，成為受歡迎的音樂製作方式之一。而 KKCompany 以「多媒體串流、數位雲端、AI 應用」為技術與事業發展核心，旗下擁有亞洲最具影響力的音樂串流品牌 —— KKBOX，對於相關趨勢與科技也非常關注。
Song generation stands out as a popular method of music production in the music creation community. KKCompany, with "multimedia technologies, digital cloud, AI applications" as the core technology and business development, owns KKBOX, the most influential music streaming brand in Asia, and is very interested in related trends and development.
近期，我們邀請到國立台灣大學電機資訊學院的楊奕軒教授，舉辦了一場「基於深度學習的自動音樂生成」內部分享會。從數據表示到 MIDI 音樂生成，延伸至音色合成以及創新的歌聲和伴奏生成應用，進行一場音樂領域的深度生成模型之旅。
Recently, we had an internal sharing session on "Deep Learning-based Automatic Music Generation" with the esteemed Professor Yi-Hsuan Yang from the College of Electrical Engineering and Computer Science, National Taiwan University! From data representations to MIDI music generation, timbre synthesis, and innovative applications in singing voice and accompaniment generation, Prof. Yang took us on a fascinating journey through the latest advancements in deep generative models for music.
除了技術分享，楊奕軒教授也與 KKCompany 核心技術研發中心 (Advanced Research Center，ARC) 的譚至斌、官順暉合力在 ACM ICMR 2024 發表最新的研究成果，介紹 Piano Cover Generation(PiCoGen)，這是一種兩階段的自動鋼琴演奏生成方法。
首先，系統將給定的歌曲轉譯為包含旋律與和弦的功能譜 (lead sheet) ，再進行鋼琴演奏生成。概念有點像是先有人聽一首流行歌，並將和弦及音符記錄下來，也就是透過 AI 抓譜；再交由另一位虛擬樂手，負責看著譜用鋼琴彈奏出來。這種方法的優勢在於，不需要翻唱和原歌曲的配對資料進行訓練即可有很好的表現。與需要配對資料的其它現有方法相比，我們的評估顯示，PiCoGen 在不同音樂類型的歌曲上表現出具有競爭力甚至更優越的性能。
為進行學術探討，團隊使用公開音樂資料庫進行研究。目前，此研究成果證明技術上已可達成鋼琴樂的自動生成，但仍需探討相關的音樂版權議題才能進一步做實際應用。
Recommended by LinkedIn
In addition to the technical sharing, Prof. Yang, together with Chih-Pin Tan and Shuen-Huei Guan from KKCompany's Advanced Research Center (ARC), published a paper in the ACM International Conference on Multimedia Retrieval (ICMR) 2024, introducing Piano Cover Generation (PiCoGen), a two-stage approach for automatic piano cover generation that transcribes the melody line and chord progression of a song given its audio recording, and then uses the resulting lead sheet as the condition to generate a piano cover in the symbolic domain. This approach is advantageous in that it does not require paired data of covers and their original songs for training. Compared to an existing approach that demands such paired data, our evaluation shows that PiCoGen demonstrates competitive or even superior performance across songs of different musical genres.
The team used the public music database for research. Current technical research indicates this method is achievable, but actual applications require relevant copyright authorization.
ICMR 為享譽盛名的國際會議，聚集了多媒體研究和技術領域的頂尖人才，一同分享突破性的創新和進展。讓我們一起往前推進多媒體檢索應用的邊界吧！
ICMR is the prestigious conference that brings together the brightest minds in multimedia research and technology to share groundbreaking innovations and advancements. Let's push the boundaries of multimedia retrieval together!
追蹤 KKCompany 掌握更多產業趨勢與資訊：