DeepRapper: Neural Rap Generation with Rhyme and Rhythm Modeling, by Lanqing Xue, Kaitao Song, Duocai Wu, Xu Tan, Nevin L. Zhang, Tao Qin, Wei-Qiang Zhang, Tie-Yan Liu, ACL 2021, is a Transformer-based rap generation system that can model both rhymes and rhythms. It generates lyrics in the reverse order with rhyme representation and constraint for rhyme enhancement and inserts a beat symbol into lyrics for rhythm/beat modeling. To our knowledge, DeepRapper is the first system to generate rap with both rhymes and rhythms.
You can specify the arguments in the bash file, such as number of epoch, bach size, etc. The trained model is saved in [model_dir]/lyrics/[raw_data_dir][_reverse]/[model_sign]/final_model. For example, in the default train.sh, the path is model/lyrics/lyrics_samples_reverse/samples/final_model.
To generate by our provided pretrained DeepRapper, first unzip the pretrained DeepRapper. Then, put the unzipped directory deeprapper-model under the folder model/. So, the complete paths are like the follows: