Looking for something that fills this gap (similar to Word embedding or Perceptual Hashing) to build a custom music recommendation/DJ tool 🤔.
There are some packages that work starting from the music notation () or MIDI files/pianorolls (github/PooyaHekmati/music_embedding/arxiv/2104.11880), but that wouldn't work as I'd like to start from the audio files (OGG, MP3, ...) and produce an embedding from there.