VieNeu-TTS-140h-nanocodec

# VieNeu-TTS-140h-nanocodec

Dataset tiếng Việt đã được encode bằng NVIDIA NeMo NanoCodec cho training Kani TTS.

## Dataset Details

- **Source:** VieNeu-TTS-140h (pnnbao-ump/VieNeu-TTS-140h)
- **Total samples:** 74,858
- **Speakers:** 193
- **Language:** Vietnamese (vi)
- **Codec:** NVIDIA NeMo NanoCodec (22kHz, 0.6kbps, 12.5fps)
- **Format:** JSONL.gz với NanoCodec tokens (4 layers)

## Dataset Structure

Mỗi mẫu bao gồm:
- `text`: Text transcription tiếng Việt
- `nano_layer_1`: Codec tokens layer 1
- `nano_layer_2`: Codec tokens layer 2
- `nano_layer_3`: Codec tokens layer 3
- `nano_layer_4`: Codec tokens layer 4
- `encoded_len`: Độ dài sequence tokens
- `speaker`: Speaker ID
- `lang`: "vi"
- `dataset_source`: "vieneu-tts-140h"
- `phonemized_text`: IPA phonemization (nếu có)
- `gender`: Giới tính speaker (nếu có)

## Usage

```python
from datasets import load_dataset

dataset = load_dataset("pnnbao-ump/VieNeu-TTS-140h-nanocodec")

# Xem một mẫu
sample = dataset['train'][0]
print(sample['text'])
print(f"Speaker: {sample['speaker']}")
print(f"Tokens: {sample['encoded_len']}")
```

## License

CC BY-NC 4.0 - Non-commercial use only

## Citation

```bibtex
@dataset{vieneu_tts_140h_nanocodec,
title = {VieNeu-TTS-140h NanoCodec Dataset},
author = {Pham Nguyen Ngoc Bao},
year = {2025},
url = {https://huggingface.co/datasets/pnnbao-ump/VieNeu-TTS-140h-nanocodec}
}
```

# VieNeu-TTS-140h-nanocodec

本数据集为越南语语音语料库，经NVIDIA NeMo NanoCodec编码后，可用于Kani TTS模型的训练。

## 数据集详情

- **来源**：VieNeu-TTS-140h（pnnbao-ump/VieNeu-TTS-140h）
- **总样本量**：74,858
- **发音人数量**：193
- **语言**：越南语（vi）
- **编解码器**：NVIDIA NeMo NanoCodec（22kHz、0.6kbps、12.5fps）
- **数据格式**：包含4层NanoCodec Token的JSONL.gz压缩文件

## 数据集结构

每个样本包含以下字段：
- `text`：越南语文本转写内容
- `nano_layer_1`：编解码器第1层Token
- `nano_layer_2`：编解码器第2层Token
- `nano_layer_3`：编解码器第3层Token
- `nano_layer_4`：编解码器第4层Token
- `encoded_len`：Token序列长度
- `speaker`：发音人ID
- `lang`：语言标识"vi"
- `dataset_source`：数据集来源"vieneu-tts-140h"
- `phonemized_text`：国际音标（IPA）音素转写结果（如存在）
- `gender`：发音人性别（如存在）

## 使用方法

python
from datasets import load_dataset

dataset = load_dataset("pnnbao-ump/VieNeu-TTS-140h-nanocodec")

# 查看一个样本
sample = dataset['train'][0]
print(sample['text'])
print(f"Speaker: {sample['speaker']}")
print(f"Tokens: {sample['encoded_len']}")

## 许可证

CC BY-NC 4.0，仅允许非商业使用。

## 引用格式

bibtex
@dataset{vieneu_tts_140h_nanocodec,
title = {VieNeu-TTS-140h NanoCodec Dataset},
author = {Pham Nguyen Ngoc Bao},
year = {2025},
url = {https://huggingface.co/datasets/pnnbao-ump/VieNeu-TTS-140h-nanocodec}
}