Results
Pianoroll Visualizations
Evolution of the generated pianorolls as a function of update steps
Randomly-chosen generated pianorolls (generated from scratch)
Audio Samples
Note that all the audio samples presented below have been downsampled to 4 time steps per beat (originally 24 time steps per beat).
Best samples
Generation from scratch
No cherry-picking. Some might sound unpleasant. Lower the volume first!
| Model | Sample |
|---|---|
| composer | |
| jamming | |
| hybrid |
Track-conditional generation
No cherry-picking. Some might sound unpleasant. Lower the volume first!
| Model | Sample |
|---|---|
| composer | |
| jamming | |
| hybrid |