README.md

---
configs:
- config_name: default
  data_files:
  - split: train
    path: data/train-*
dataset_info:
  features:
  - name: source_dataset
    dtype: string
  - name: scene_id
    dtype: string
  - name: char_id
    dtype: string
  - name: audio
    dtype: audio
  - name: text
    dtype: string
  - name: text_source
    dtype: string
  - name: duration
    dtype: float64
  - name: sampling_rate
    dtype: int64
  - name: asr_model
    dtype: string
  splits:
  - name: train
    num_bytes: 259702003422
    num_examples: 1033142
  download_size: 240712376248
  dataset_size: 259702003422
license: mit
task_categories:
- text-to-speech
- automatic-speech-recognition
language:
- ja
tags:
- speech
- audio
- voice
- japanese
- anime
pretty_name: Japanese-Eroge-Voice-V2
size_categories:
- 1M<n<10M
---

# Japanese-Eroge-Voice-V2

## Description

This is the successor to the **[Japanese-Eroge-Voice](NandemoGHS/Japanese-Eroge-Voice)** dataset. It consists of a significantly larger collection of audio-transcription pairs extracted from Japanese eroge (adult games). 

**Note on Versioning:** There is **no overlap** between this dataset (V2) and the previous version. All audio clips and transcriptions in V2 are distinct from those in the original version, providing entirely new data for research.

This version (V2) expands the total audio duration to over 2,600 hours, providing a more robust resource for speech synthesis and recognition research.

-----

## Dataset Statistics

- **Total Clips:** 1,033,142
- **Total Duration:** Approx. 2,657 hours 6 minutes 22 seconds
- **Mean Audio Length:** 9.259 seconds
- **Median Audio Length:** 8.893 seconds

-----

## Key Columns:
- **`audio`**: The audio data.
- **`text`**: The text corresponding to the audio.
- **`text_source`**: Indicates the origin of the transcription:
  - `original`: Transcriptions parsed directly from the game's original script files.
  - `asr`: Transcriptions generated using the **[litagin/anime-whisper](https://huggingface.co/litagin/anime-whisper)** model.
- **`sampling_rate`**: The original sampling rate of the audio clip.
- **`source_dataset`, `scene_id`, `char_id`**: Anonymized via hashing.

-----

## Anonymization and Shuffling

To comply with Japanese copyright law and protect the integrity of the original works:
1. All metadata identifiers (`source_dataset`, `scene_id`, `char_id`) have been **hashed**.
2. The dataset has been **randomly shuffled**.

These steps ensure that it is practically impossible to reconstruct the original narratives or games in their entirety, aiming to limit the use of the data to research and educational purposes rather than the consumption of copyrighted content.

-----

## Biases and Limitations

* **NSFW Content**: This dataset is derived from adult games and contains a significant amount of Not-Safe-For-Work (NSFW) content.
* **Gender Bias**: Due to the nature of the source material, the dataset is heavily skewed towards female voices.
* **Potential Transcription Errors**: While some transcriptions are from original scripts, others are generated by AI (`asr`) and have not been manually verified. These may contain inaccuracies.

-----

## License

This dataset is licensed under the **[MIT License](https://choosealicense.com/licenses/mit/)**.

**Intended use** — This dataset is primarily designed for **educational and academic research. All use is at your own risk, and you must ensure compliance with applicable law.**

**NO WARRANTY** – This dataset is provided “as is” without any express or implied warranty.

## How to Cite

```bibtex
@misc{Japanese-Eroge-Voice-V2,
  title        = {Japanese-Eroge-Voice-V2},
  author       = {OmniAICreator},
  year         = {2025},
  howpublished = {\url{https://huggingface.co/datasets/NandemoGHS/Japanese-Eroge-Voice-V2}},
}