PianoKontext: Expressive Performance Rendering from Deadpan Context

PianoKontext is a flow matching rendering model for classical piano music that generates variable-length performances in the latent space of a pretrained Music2Latent model.

• Paper: PianoKontext: Expressive Performance Rendering from Deadpan Context
• GitHub Repository: realfolkcode/pianokontext
• Project Page & Demos: PianoKontext Demo
• Interactive Demo: Open In Colab

Description

Expressive performance rendering (EPR) aims to generate realistic performances constrained on sequences of notes. PianoKontext addresses the limitations of existing flow matching models regarding expressive timing by generating variable-length performances. It synthesizes MIDI scores into deadpan audio and employs Dynamic Time Warping (DTW) in the latent space to construct paired data for training. The aligned embeddings are concatenated in DiT blocks, allowing the model to learn the dependencies between the score and expressive performances.

Citation

If you use this model in your research, please cite the following paper:

@article{gavrilev2026pianokontext,
  title={PianoKontext: Expressive Performance Rendering from Deadpan Context},
  author={Dmitrii Gavrilev},
  journal={arXiv preprint arXiv:2606.12282},
  year={2026}
}