---
license: cc-by-4.0
configs:
  - config_name: default
    drop_labels: true
---

# VocalGrad Representative

This repository contains a compact representative subset of VocalGrad for quick
inspection and smoke testing.

It contains one base source clip and all augmentations derived from that clip
for the nine active VocalGrad categories.

## Contents

- source clip: `p226_106`
- source text: "It was some time before she found out he was safe."
- categories: `9`
- augmentations per category: `24`
- total augmented clips: `216`
- labels: `increase`, `decrease`

## Layout

```text
README.md
metadata.csv
audio/
source_audio/
source_text/
metadata/summary.json
```

The main data file follows the Hugging Face audio dataset convention:

- `metadata.csv` is located at the repository root.
- `metadata.csv` contains a `file_name` column.
- `file_name` contains full relative paths to the benchmark clips in the flat
  `audio/` directory.

## Task

Given an audio clip and an attribute name, answer whether the target attribute
increases or decreases over time.

```text
Does the {attribute} increase or decrease over time?

Answer with only one word: "increase" or "decrease".
```

The label is derived from `direction`:

- `up` -> `increase`
- `down` -> `decrease`

## Categories

- `volume`
- `speaking_speed`
- `voice_pitch`
- `background_noise`
- `audio_distortion`
- `audio_roughness`
- `voice_clarity`
- `voice_vibration`
- `echo`

## License

This representative VocalGrad subset is released under the Creative Commons
Attribution 4.0 International License (CC BY 4.0).

The source speech clip is derived from the CSTR VCTK Corpus version 0.92, which
is distributed under the Creative Commons Attribution 4.0 International Public
License.

## Source Dataset Credit

VocalGrad uses source speech from the CSTR VCTK Corpus:

Yamagishi, Junichi; Veaux, Christophe; MacDonald, Kirsten. (2019). CSTR VCTK
Corpus: English Multi-speaker Corpus for CSTR Voice Cloning Toolkit (version
0.92), [sound]. University of Edinburgh. The Centre for Speech Technology
Research (CSTR). https://doi.org/10.7488/ds/2645.

Original dataset record:
https://datashare.ed.ac.uk/handle/10283/3443

License:
Creative Commons Attribution 4.0 International
https://creativecommons.org/licenses/by/4.0/

## Modifications

The audio clips in this representative subset are transformed derivatives of
one VCTK source clip. VocalGrad applies controlled time-varying transformations
to create benchmark examples for detecting whether a target perceptual
attribute increases or decreases over time. The included metadata identifies
the source clip ID and the transformation condition for each derived clip.