HARP is a neural audio codec that partitions residual vector quantization across harmonically meaningful frequency bands. A harmonic-aware partitioning distributes codebook capacity over perceptually important bands, so a single model serves multiple bitrates by decoding a growing number of codebook groups. This page presents audio reconstructions across HARP's bitrate tiers.
Audio samples
Each sample shows the reference recording and HARP vs. the DAC baseline at 4.3 and 7.7 kbps.
No samples yet. Generate them with
python scripts/build_demo.py your_audio/*.wav,
then commit docs/audio/.
Bitrate tiers
| Groups | Codebooks | Bitrate |
|---|---|---|
| 1 | 3 | ~2.6 kbps |
| 2 | 5 | ~4.3 kbps |
| 3 | 7 | ~6.0 kbps |
| 4 | 9 | ~7.7 kbps (full) |
Citation
@inproceedings{harp2026,
title = {HARP: Harmonic-Aware Residual Partitioning for Neural Audio Codecs},
author = {Yang, Qiaoyu and He, Lixing and Deng, Binyue and Zhao, Weifeng},
booktitle = {Interspeech},
year = {2026}
}