---
license: other
language:
- en
task_categories:
- text-to-speech
- automatic-speech-recognition
- audio-classification
pretty_name: NaturalVoices Restored (16 kHz, Sidon + UTMOS-filtered)
size_categories:
- 100K<n<1M
tags:
- speech
- audio
- voice
- speech-restoration
- sidon
- utmos
- english
configs:
- config_name: default
  data_files:
  - split: train
    path: data/train-*
dataset_info:
  features:
  - name: file_name
    dtype: string
  - name: Document
    dtype: string
  - name: Part Number
    dtype: int64
  - name: DNSMOSPro
    dtype: float64
  - name: gender
    dtype: string
  - name: ASR_CONF
    dtype: float64
  - name: SNR
    dtype: float64
  - name: NUM_SPKS
    dtype: int64
  - name: Arousal
    dtype: float64
  - name: Dominance
    dtype: float64
  - name: Valence
    dtype: float64
  - name: Neutral
    dtype: float64
  - name: Angry
    dtype: float64
  - name: Sad
    dtype: float64
  - name: Happy
    dtype: float64
  - name: Emotion
    dtype: string
  - name: total time
    dtype: float64
  - name: start
    dtype: float64
  - name: end
    dtype: float64
  - name: text
    dtype: string
  - name: utmos
    dtype: float64
  - name: utmos_orig
    dtype: float64
  - name: age
    dtype: float64
  - name: wpm
    dtype: float64
  - name: audio
    dtype:
      audio:
        sampling_rate: 16000
  splits:
  - name: train
    num_bytes: 202098673799.464
    num_examples: 494903
  download_size: 169904567123
  dataset_size: 202098673799.464
---

# NaturalVoices — Sidon-Restored, UTMOS-Filtered (16 kHz)

High-quality English speech derived from **[NaturalVoices_VC_870h](https://huggingface.co/datasets/JHU-SmileLab/NaturalVoices_VC_870h)** (JHU SmileLab), restored
with **[Sidon v0.1](https://huggingface.co/sarulab-speech/sidon-v0.1)** and kept **only**
where restoration measurably improved perceptual quality (UTMOS gate). Each clip ships
with rich per-utterance metadata (transcript, speaker age/gender, speaking rate, emotion,
and quality scores) so it is ready for TTS / voice-cloning / ASR / paralinguistic research.

- **494,903 clips · 736.7 hours** (clips ≥ 3.0 s)
- **16 kHz**, mono, float32, embedded audio (plays in the HF viewer)
- 48 kHz twin: [`PleasedPenguin/naturalvoice_737h_48k`](https://huggingface.co/datasets/PleasedPenguin/naturalvoice_737h_48k) — *same utterances, different sampling rate*

## Quick start

```python
from datasets import load_dataset

ds = load_dataset("PleasedPenguin/naturalvoice_737h_16k", split="train")
ex = ds[0]
print(ex["text"], ex["age"], ex["gender"], ex["wpm"], ex["utmos"])
audio = ex["audio"]            # {'array': np.ndarray, 'sampling_rate': 16000}
```

## How it was built

1. **Restore** — every source clip is passed through Sidon v0.1 (deterministic feed-forward
   restoration, no sampling) to a 48 kHz waveform; the 16 kHz version is a clean downsample
   of that same restored signal.
2. **Quality gate** — UTMOS (`utmos22_strong`) is scored on the original and the restored
   clip. A clip is **kept only if `utmos > utmos_orig`** (restoration actually helped).
   Both scores are stored so you can re-threshold.
3. **Label** — speaker **age** (audeering `wav2vec2-large-robust-24-ft-age-gender`) and
   **wpm / speaking rate** (Qwen3 Forced Aligner 0.6B: aligned word count over the
   first-word-onset → last-word-offset span) are added. The original NaturalVoices
   annotations (emotion, SNR, DNSMOS, etc.) are carried through unchanged.

## Columns

| Column | Description |
|---|---|
| `audio` | Sidon-restored waveform (16 kHz, float32) |
| `text` | Transcript |
| `file_name`, `Document`, `Part Number` | Clip identifiers |
| `total time`, `start`, `end` | Clip duration / source offsets (seconds) |
| `gender` | Speaker gender (original NaturalVoices label) |
| `age` | Predicted speaker age in years (audeering wav2vec2) |
| `wpm` | Speaking rate (words per minute), forced-aligned |
| `utmos` | UTMOS of the **restored** clip |
| `utmos_orig` | UTMOS of the **original** clip (kept only when `utmos > utmos_orig`) |
| `DNSMOSPro`, `SNR`, `ASR_CONF`, `NUM_SPKS` | Original NaturalVoices quality annotations |
| `Arousal`, `Dominance`, `Valence` | Original emotion attributes |
| `Neutral`, `Angry`, `Sad`, `Happy`, `Emotion` | Original emotion labels |

## Dataset statistics

**Speaking rate (wpm)** — mean **175**, median **171**, p5–p95 **104–261**.

**Age** — mean **36.5 yr**, median **33.7**, p10–p90 **22.8–55.4**.

**Gender** — model vs. original-label agreement **94.2%** (the original `gender` label is
the one shipped in this dataset; the predictor was used only for cross-checking).

## Source & license

Derived from **[NaturalVoices_VC_870h](https://huggingface.co/datasets/JHU-SmileLab/NaturalVoices_VC_870h)**
(JHU SmileLab). Audio has been restored/transformed; please honor the terms of the original
NaturalVoices release for any redistribution or use. Restoration: Sidon v0.1.
Quality metric: UTMOS (`tarepan/SpeechMOS`).

## Citation

If you use this data, please cite the original NaturalVoices dataset:

```bibtex
@misc{du2025naturalvoiceslargescalespontaneousemotional,
      title={NaturalVoices: A Large-Scale, Spontaneous and Emotional Podcast Dataset for Voice Conversion},
      author={Zongyang Du and Shreeram Suresh Chandra and Ismail Rasim Ulgen and Aurosweta Mahapatra and Ali N. Salman and Carlos Busso and Berrak Sisman},
      year={2025},
      eprint={2511.00256},
      archivePrefix={arXiv},
      primaryClass={eess.AS},
      url={https://arxiv.org/abs/2511.00256},
}
```