---
license: mit
library_name: pytorch
pipeline_tag: audio-to-audio
tags:
  - audio
  - audio-restoration
  - audio-super-resolution
  - speech-enhancement
  - music-enhancement
  - pytorch
  - aero
---

# AERO Audio Super-Resolution (Multi-Variant Checkpoint Pack)

This repository is a reupload of AERO weights for _AERO: Audio Super Resolution in the Spectral Domain_.

- [github](https://github.com/slp-rl/aero)
- [paper](https://arxiv.org/abs/2211.12232)
- [project page](https://pages.cs.huji.ac.il/adiyoss-lab/aero/)
- [original weights](https://drive.google.com/drive/folders/1KuVJNkR7lZddvufmNsx-uAIluvb5XQ2L)

## Training Data

Original AERO training setup includes:

- Music: MUSDB-style data preparation for band extension.
- Speech: VCTK-derived low/high sample-rate pairs (4-16, 8-16, 8-24, 12-48).