Topic
A new piano dataset pairs 18,786 performances with their written scores across pop, classical, jazz, folk, rock and blues. It could help transcription and humanising tools move beyond classical music, but the data is non-commercial only and the audio itself is not redistributed.
AIvia arXiv (Zeng, Mathur, Wang)
Audio vs MIDI is sound vs instructions. Audio is a recording of sound waves: what you hear is fixed in the file. MIDI is a list of notes and how to play them, with no sound until an instrument plays it, so every note can be changed later. That difference decides how you edit, what you can export, and which AI music tools hand you something you can actually work on.
Workflows
A free chain for getting from a Suno-style song to parts a live band can read: split the audio into stems, convert each stem to MIDI with Spotify's Basic Pitch, clean it up and print parts in MuseScore, then write a one-page chord chart by ear. Every tool here is free; the time goes into the clean-up.
Workflows
One email a week: the music-tech and AI-in-music news that mattered, and what we think about it.