Without this dataset, you can’t train a supervised riff generator. Mandatory reading. 6. Practical Implementation Reference (Code + Paper) Title: MuseGuitar: Guitar Tablature Generation with Attention and Pitch-Shift Augmentation Venue: ICASSP 2023 workshop Key Idea: Encoder-decoder with pitch-shift data augmentation to avoid overfitting to key signatures.

You can condition generation on “blues riff” vs “thrash metal riff” – very practical for a generator. 4. Interactive & Real-Time Riff Generation Title: RiffLoop: A Loop-Based Guitar Riff Generator Authors: B. Hayes, C. Saitis (NIME 2021) Key Idea: Generates 1- or 2-bar riffs that can be looped, with a user-adjustable “groove” parameter.

It looks like you're looking for a paper (or a set of useful references) on , likely in the context of music AI, LSTMs, transformers, or symbolic music generation.