Benchmarking neural music editing with controlled MIDI changes
SMEBENCH: Benchmarking Neural Music Editing with Synthetized MIDI Edits
SoundMachine Learning
Summary
Editing music with AI is tricky because changes must be made without messing up the rest of the song. The authors created SMEBench, a big set of music examples where specific changes to MIDI files are paired with the correct resulting audio. This setup allows precise testing of how well editing systems keep the original sound and apply the requested changes. They tested six editing tools and found that choosing how to measure success can change which tool seems best. The dataset helps researchers and developers better evaluate AI music editors.
What this means in practice
- •For music software developers: Evaluate and improve AI-powered music editing tools with precise paired MIDI and audio examples enabling clear measurement of edit quality.
- •For audio engineers: Assess the effectiveness of automated music editing systems in tasks like changing instrument sounds or dynamics while preserving original content.
Authors
Matthieu Cervera, Francesco Paissan, Micael Antunes, Mirco Ravanelli, Ricard Marxer, Cem Subakan
Abstract
Neural music editing requires a system to perform a requested change while preserving the musical content that should remain unchanged. Existing evaluations often lack a paired target edit, making these requirements difficult to measure separately. We introduce SMEBench, a large-scale benchmark of controlled music edits constructed by modifying symbolic MIDI performances and rendering matched source and target audio. Each example contains a source, an instruction, a paired target, edit metadata, and complementary descriptions. The benchmark covers twelve instrumentation, articulation, dynamics, and pitch operations and includes a balanced 3,600-example test set together with a smaller real-audio complement derived from multitrack recordings. The paired targets allow source preservation, target agreement, and textual compatibility to be evaluated independently. We use these references to diagnose common audio-editing metrics and benchmark six representative editing systems, showing that the metric used to rank a system can materially change the conclusion. The dataset is available at our repository.