---
language:
- ru
license: apache-2.0
task_categories:
- text-to-speech
- automatic-speech-recognition
tags:
- audio
- speech
- russian
- tts
size_categories:
- 10K<n<100K
configs:
- config_name: default
  data_files:
  - split: train
    path: data/train-*
dataset_info:
  features:
  - name: audio
    dtype:
      audio:
        sampling_rate: 16000
  - name: text
    dtype: string
  splits:
  - name: train
    num_bytes: 7479621859.359
    num_examples: 15813
  download_size: 7197154270
  dataset_size: 7479621859.359
---

# ToneBooksPlus-Grigorii

Filtered subset of [ToneBooksPlus](https://huggingface.co/datasets/Vikhrmodels/ToneBooksPlus) containing only samples from speaker **Grigorii Metelitsa**.

## Dataset Details

- **Speaker:** Grigorii Metelitsa (professional Russian voice actor)
- **Total Samples:** 15,813
- **Total Duration:** ~29 hours
- **Sample Rate:** 16,000 Hz
- **Format:** WAV (mono, 16-bit PCM)
- **Language:** Russian

## Source

This dataset is a filtered subset of the original ToneBooksPlus dataset:
- **Original Dataset:** [Vikhrmodels/ToneBooksPlus](https://huggingface.co/datasets/Vikhrmodels/ToneBooksPlus)
- **Filtering Criteria:** Only samples from speaker "grigorii_metelitsa"
- **License:** Apache 2.0 (inherited from ToneBooksPlus)

## Dataset Structure

Each sample contains:
- `audio`: Audio file (WAV, 16kHz, mono)
- `text`: Transcribed Russian text

### Example

```python
from datasets import load_dataset

dataset = load_dataset("ALEKAS/ToneBooksPlus-Grigorii")

# Access first sample
sample = dataset['train'][0]
print(sample['text'])
# Audio is automatically loaded as array
audio_array = sample['audio']['array']
sample_rate = sample['audio']['sampling_rate']
```

## Use Cases

- Russian TTS model training
- Voice cloning (Grigorii's voice)
- Russian ASR model fine-tuning
- Speech synthesis research

## Citation

If you use this dataset, please cite the original ToneBooksPlus:

```bibtex
@misc{tonebooksplus,
  title={ToneBooksPlus: Russian Audiobook Dataset},
  author={Vikhrmodels},
  year={2024},
  publisher={Hugging Face},
  howpublished={\url{https://huggingface.co/datasets/Vikhrmodels/ToneBooksPlus}}
}
```

## License

Apache 2.0 (same as original ToneBooksPlus dataset)

## Acknowledgements

- Original dataset creators: Vikhrmodels
- Voice actor: Grigorii Metelitsa
- Dataset curation: ALEKAS