AI Empowered Audio-to-MIDI conversion

AI 赋能音频到 MIDI 转换

• 批准号: 10063393
• 金额: $ 6.33万
• 依托单位: ESPRESSIVO LTD.
• 依托单位国家: 英国
• 项目类别: Collaborative R&D
• 财政年份: 2023
• 资助国家: 英国
• 起止时间: 2023 至 无数据
• 项目状态: 已结题
• 来源: https://gtr.ukri.org/projects?ref=10063393
• 关键词: AI; Empowered; Audio; MIDI; conversion

Music creators who prefer to compose by playing an acoustic instrument instead of using music notation or a MIDI instrument with a digital audio workstation (DAW) often encounter challenges in converting their audio recordings into MIDI data. The current technologies for Audio-to-MIDI conversion are often unreliable, particularly for complex musical compositions, and the process of correcting errors can be both time-consuming and tedious. This difficulty prevents professional and amateur composers, producers, and music educators from taking full advantage of the many tools and resources offered by the music production software industry. For example, many software packages feature automated accompaniment and harmonization tools, which are made far less useful if the audio track is not converted to a MIDI format.We have developed a self-correcting machine learning application that can accurately convert Audio to MIDI, regardless of the complexity, tempo, or instrumentation used. The goal of our proposed project is to make our technology available to all by creating a professional plugin.

喜欢通过演奏声学乐器而不是使用乐谱或数字音频工作站（Digital Audio Workstation，MIDI）的乐器来作曲的音乐创作者通常会在将其音频记录转换为MIDI数据时遇到挑战。目前的音频到音频转换技术通常不可靠，特别是对于复杂的音乐作品，纠正错误的过程既耗时又繁琐。这种困难阻碍了专业和业余作曲家、制作人和音乐教育工作者充分利用音乐制作软件行业提供的许多工具和资源。例如，许多软件包都具有自动伴奏和和声工具，如果音轨没有转换为MIDI格式，这些工具就变得不那么有用了。我们开发了一个自校正机器学习应用程序，可以准确地将Audio转换为MIDI，无论使用的复杂性、克里思或乐器如何。我们提出的项目的目标是通过创建一个专业的插件，使我们的技术提供给所有人。