Includes a GitHub repo with pretrained models for riff generation. Summary Recommendation Table | If you want… | Start with… | |--------------|--------------| | Academic foundation | Generating Guitar Riffs with Deep Learning (Cassotti) | | Rock/metal style | JamBot (Donahue) | | Control over genre/style | GTR-CTRL (Sarmento) | | Loop-based / DAW-friendly | RiffLoop (Hayes) | | Dataset & preprocessing | GuitarSet (Xi) | | Working code to run | MuseGuitar (ICASSP 2023) |
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. guitar riff generator
Handles riff repetition with variation – a key property of riffs (not just random notes). 3. Transformer-Based Approach (State-of-the-Art) Title: GTR-CTRL: Instrument and Genre Conditioning for Guitar Tablature Generation Authors: P. Sarmento, et al. (2022, arXiv) Key Idea: Uses a Transformer XL model trained on thousands of Guitar Pro files, with conditioning on tuning, genre, and tempo. Includes a GitHub repo with pretrained models for