OpenVoiceOS/tts-vc-mcv-scripted-v24.0-fy-nl-dii

---
language:
- fy
license: cc0-1.0
library_name: datasets
tags:
- audio
- text-to-speech
- tts
- frisian
- mozilla-common-voice
datasets:
- mozilla-foundation/common_voice_17_0
metrics:
- wer
---

# OVOS - TTS Voice-Converted Mozilla Common Voice Scripted Speech v24.0 - Frisian Dataset

This dataset is a refined, processed version of the **Mozilla Common Voice (MCV) 24.0** Frisian (`fy-NL`) corpus, specifically curated for **Text-to-Speech (TTS)** training.

Unlike the raw ASR-focused Common Voice data, this version has been filtered for high-quality audio, normalized text, and formatted for modern TTS architectures like VITS, Glow-TTS, or YourTTS.

## Dataset Description
- **Language:** Frisian (fy-NL)
- **Source:** [Mozilla Common Voice Scripted Speech v24.0 - Frisian](https://datacollective.mozillafoundation.org/datasets/cmj8u3p4d008tnxxbxw3bkw70)
- **Format:** 24,000Hz Mono WAV
- **License:** [CC0 1.0 Universal](https://creativecommons.org/publicdomain/zero/1.0/)

## Processing Steps
To make this data suitable for TTS, the following pipeline was applied:
1. **Format conversion:** Conversion from mp3 to wav.
2. **Denoising:** Suppression of background noise and correction of distortion.
3. **Trimming:** Stripped silence from all clips.
4. **Normalization:** All audio intensity normalized.
5. **Outlier filtering:** Outliers filtered out in terms of unusual words per minute.
6. **Voice conversion:** Conversion of speaker features in the audio to a donated voice (to OVOS).

## Dataset Structure
The dataset follows a standard metadata format:
* `file_name`: The audio file path or array.
* `transcript`: The normalized transcript for TTS.

An alternative metadata file is provided with the original Mozilla Common Voice metadata.

## Licensing
Consistent with the [Mozilla Common Voice Legal Terms](https://commonvoice.mozilla.org/en/terms), this dataset is released under the **Creative Commons CC0 1.0 Universal (Public Domain)** license. You may use, modify, and distribute this data without restriction.

## Citation
If you use this dataset, please cite the original Common Voice project:
```bibtex
@inproceedings{commonvoice:2020,
author = {Ardila, R. and others},
title = {Common Voice: A Massively-Multilingual Speech Corpus},
booktitle = {Proceedings of the 12th Conference on Language Resources and Evaluation (LREC 2020)},
pages = {4211--4215},
year = {2020}
}

---
语言：
- fy
许可证：CC0 1.0
库名称：datasets
标签：
- 音频
- 文本到语音（Text-to-Speech）
- TTS
- 弗里斯兰语
- Mozilla Common Voice
数据集：
- mozilla-foundation/common_voice_17_0
指标：
- 词错误率（Word Error Rate, WER）
---

# OVOS——TTS语音转换型Mozilla Common Voice脚本语音v24.0——弗里斯兰语数据集

本数据集为经过优化与处理的**Mozilla Common Voice (MCV) 24.0**弗里斯兰语（fy-NL）语料库版本，专为**文本到语音（Text-to-Speech, TTS）**训练精心打造。

与专注于自动语音识别（Automatic Speech Recognition, ASR）的原始Common Voice数据不同，本版本经过筛选以保留高质量音频、归一化文本，并针对VITS、Glow-TTS、YourTTS等现代TTS架构完成格式适配。

## 数据集描述
- **语言：** 弗里斯兰语（fy-NL）
- **来源：** [Mozilla Common Voice Scripted Speech v24.0 - 弗里斯兰语](https://datacollective.mozillafoundation.org/datasets/cmj8u3p4d008tnxxbxw3bkw70)
- **格式：** 24000Hz单声道WAV
- **许可证：** [CC0 1.0 通用协议](https://creativecommons.org/publicdomain/zero/1.0/)

## 处理流程
为使数据适配TTS任务，我们采用了如下处理流水线：
1. **格式转换：** 将音频格式从mp3转换为wav。
2. **降噪处理：** 抑制背景噪声并修正音频失真。
3. **静音修剪：** 移除所有音频片段中的静音片段。
4. **响度归一化：** 统一所有音频的响度水平。
5. **异常值过滤：** 过滤掉每分钟单词数异常的音频片段。
6. **语音转换：** 将音频中的说话人特征转换为捐赠给OVOS的目标语音。

## 数据集结构
本数据集采用标准元数据格式：
* `file_name`：音频文件路径或数组。
* `transcript`：用于TTS任务的归一化文本转录结果。

同时提供包含原始Mozilla Common Voice元数据的备用元数据文件。

## 许可证说明
遵循[Mozilla Common Voice法律条款](https://commonvoice.mozilla.org/en/terms)，本数据集采用**知识共享CC0 1.0通用（公共领域）**许可证发布。您可无限制地使用、修改和分发本数据集。

## 引用方式
若您使用本数据集，请引用原始Common Voice项目：
bibtex
@inproceedings{commonvoice:2020,
author = {Ardila, R. and others},
title = {Common Voice: A Massively-Multilingual Speech Corpus},
booktitle = {Proceedings of the 12th Conference on Language Resources and Evaluation (LREC 2020)},
pages = {4211--4215},
year = {2020}
}