Decomposer: Decompile Symbolic Music to Programs (github.com)

A new tool called Decomposer has been introduced, which transforms symbolic music files (MIDI) into executable and editable Strudel programs. By leveraging a post-trained model, Decomposer allows users to decompile MIDI files into Strudel code, enhancing the ability to manipulate music programmatically. This tool provides two inference backends — a simpler, slower option (generate_transformers.py) and a faster, more complex option (generate_vllm.py) — catering to various user needs in terms of performance and setup. The significance of Decomposer lies in its potential to democratize music programming and streamline creative processes for musicians and developers alike. With features like instrument filtering, customizable duration, and sampling parameters, it opens new avenues for interactive music composition and experimentation. This tool is particularly beneficial for those working with short clips of music, as it thrives on MIDI inputs under 30 seconds. As the field of AI-driven music software continues to evolve, Decomposer represents a notable advancement, enabling broader accessibility and innovation in music generation and manipulation within the AI/ML landscape.