Apolo: a toolkit for generative music production and benchmarking

阿波罗项目是对人工智能和机器学习最新发展的直接回应，当应用于创造性任务的部分或完全自动化时，这一领域被称为计算创造力，创造性AI或元创造。特别是，基于语料库的生成音乐算法的最新进展正在导致专业人士和业余爱好者如何创作音乐的革命。这些越来越强大的算法尚未得到广泛应用，主要是由于机器学习的复杂性，以及缺乏开源库，数据管理软件和图形用户界面来访问这些算法。Apolo项目旨在填补这一空白，提供用户友好的技术，使作曲家，声音设计师，游戏开发人员，VR工作室，业余音乐家以及研究人员和学生能够在自己的应用或研究中理解，应用，改进，基准测试，扩展这些新的计算技术。阿波罗项目的主要成果将是：*-使用新奇和惊喜搜索与深度学习生成算法相结合，超越模仿和风格模仿的生成算法 *-高级变量和跨模态映射策略，可用于在交互式环境中引导音乐生成，无论是数字音频工作站中的计算机辅助作曲，或在线应用程序（如视频游戏）中的自适应在线生成。*- ApoloC语料库管理器将允许用户（研究人员，学生，音乐家，作曲家和爱好者）加载和操作音乐作品（以各种格式），以管理生成风格模仿算法的训练集。ApoloG生成引擎将提供一个图形界面来选择训练集、训练算法，并指定生成任务和参数。ApoloC和ApoloG都将依赖于Apolo库：一个开源库，实现了现有和新的音乐生成算法的集合。*-用于风格模仿的ApoloQ定量评估工具箱将被集成到框架中，以帮助基准算法。随着从线性媒体向非线性媒体的转变，用户现在可以接触到丰富的沉浸式和交互式体验。无论是游戏、教育还是娱乐，互动媒体都在迅速超越线性媒体（书籍、电影、电视等）。随着非线性（交互式）媒体的发展，所需的资产数量呈指数级增长。除了对人工智能，机器学习，计算创造力，计算机音乐和音乐信息检索的贡献外，Apolo项目还将为计算机辅助或完全程序化的音乐生成方法奠定基础，这些方法现在需要各种软件产品（交互式系统，视频游戏等），在这些行业中，数字经济和加拿大工业的核心。

The Apolo project is a direct response to recent developments in Artificial Intelligence, and Machine Learning when applied to the partial or complete automation of creative tasks, a field known as computational creativity, Creative AI, or Metacreation. In particular, recent advances in corpus-based generative music algorithms are leading to a revolution in how music is being created by professionals and amateurs alike. These increasingly powerful algorithms have not yet been applied much, mostly due to the complexities of machine learning, and the lack of an open source library, data management software, and graphical user interfaces for accessing these algorithms. The Apolo project aims to fill this gap providing user-friendly technologies that enable composers, sound designers, game developers, VR studios, amateur musicians as well as researchers and students to understand, apply, refine, benchmark, extend these new computational techniques in their own applications or research.***The main outcomes of the Apolo project will be: ***- Generative algorithms that go beyond pastiche and style imitation using novelty and surprise search in combination with Deep Learning generative algorithms***- High-level variables and cross-modal mapping strategies that can be used to steer the musical generation in interactive contexts be it for computer-assisted composition in digital audio workstations, or adaptive online generation in online applications (such as video games).***- The ApoloC corpus manager will allow users (researchers, students, musicians, composer, and enthusiasts) to load and manipulate music compositions (in a variety of formats) to manage training sets for generative style imitation algorithms.***- The ApoloG generation engine, will provide a graphic interface to select training sets, training algorithms, and specify the generative task and parameters.***- Both ApoloC and ApoloG will rest on the Apolo library: an open source library implementing a collection of existing and new music generation algorithms. ***- The ApoloQ quantitative evaluation toolbox for style imitation will be integrated into the framework to help benchmark algorithms.***With the shift from linear to non-linear media, users are now exposed to rich immersive and interactive experiences. Be it for gaming, education, or entertainment, interactive mediums are quickly outgrowing linear media (books, film, tv, ...). As nonlinear (interactive) mediums develop, the number of assets needed is growing exponentially.***Besides its contribution to artificial intelligence, machine learning, computational creativity, computer music, and music information retrieval, the Apolo project will lay the ground for computer-assisted or completely procedural music generation approaches that are now required in a variety of software products (interactive systems, video games, ...), central to the digital economy and Canadian industry in these sectors.**