Music AI

Automatic Music Transcription (AMT) has ad- vanced significantly for the piano, but transcrip- tion for the guitar remains limited due to several key challenges. Existing systems fail to detect and annotate expressive techniques (e.g., slides, bends, percussive hits) and map notes to the in- correct string and fret combination in the gen- erated tablature. Furthermore, prior models are typically trained on professionally recorded, iso- lated datasets, limiting their generalizability to varied acoustic environments with background noise such as home recordings made on standard smartphones. To overcome these limitations, we propose TART, a four-stage end-to-end pipeline that produces detailed guitar tablature directly from guitar audio. Our system consists of (1) a CRNN-based audio-to-MIDI transcription model; (2) a CNN-BiLSTM for expressive technique clas- sification; (3) a Transformer-based string and fret assignment model; and (4) an automated tablature generator, all consolidated into a pipeline that can output tablature from a given audio sample. To the best of our knowledge, this framework is the first to generate detailed tablature sheet music with accurate fingerings and expressive technique labels from guitar audio.