Recently, an open-source music generation model called NotaGen has emerged, quickly becoming a hot topic in the intersection of AI technology and art. Its core strength lies in generating high-quality classical music, but it also supports the creation of pop music, showcasing AI's powerful potential in the music field.
NotaGen employs a training paradigm similar to large language models (LLMs), pre-trained on a massive dataset of over 1.6 million musical works, ensuring its generation capabilities reach an astonishing level.
Yesterday, the developers announced the full open-sourcing of NotaGen's model and code, offering free access to music enthusiasts, developers, and researchers worldwide. This release includes not only the base version but also an enhanced model, NotaGen-X, inspired by DeepSeekR1. NotaGen-X requires 24GB of VRAM for local deployment and offers significantly improved performance, ideal for users seeking the best possible experience. However, it's important to note that NotaGen doesn't generate directly playable audio files; instead, it outputs symbolic musical scores in ABC and XML formats for easy editing and conversion.
NotaGen's core advantage lies in its high degree of controllability and professionalism. Users can specify the musical period style, such as Baroque, Classical, or Romantic, and choose specific instrument types, like keyboard or orchestral instruments, to generate tailored scores. To further enhance the quality of classical music generation, the model was fine-tuned on a specialized dataset of 8,948 classical scores encompassing 152 composers. Test results show that the generated scores are close to professional composition standards, rich in detail and stylistically accurate.
Open-sourcing significantly broadens NotaGen's applications. For professional composers, it's an efficient inspiration tool; for amateurs, it lowers the barrier to music creation. Commentators have noted that NotaGen's method of generating scores using "period-composer-instrument" prompts is not only simple to operate but also caters to diverse creative needs. Furthermore, its performance in the pop music domain is highly anticipated, and future expansion to support more musical styles is expected.
The release of NotaGen marks a new stage for AI music generation technology. It not only injects technological vitality into artistic creation but also promotes community collaboration and innovation through its open-source model. Whether it's the preservation of classical music or the exploration of pop music, NotaGen showcases limitless possibilities and deserves continued attention.