README.md

---
dataset_info:
  features:
  - name: audio
    dtype: audio
  - name: text
    dtype: string
  - name: age
    dtype: string
  - name: gender
    dtype: string
  - name: client_id
    dtype: string
  splits:
  - name: generated
    num_bytes: 30066361717.688
    num_examples: 137419
  download_size: 27873351423
  dataset_size: 30066361717.688
configs:
- config_name: default
  data_files:
  - split: generated
    path: data/generated-*
license: cc0-1.0
language:
- hy
---

# Hy-Generated Audio Data 2

This dataset provides Armenian speech data consisting of generated audio clips and is addition to [this](https://huggingface.co/datasets/ErikMkrtchyan/Hy-Generated-audio-data-with-cv20.0) dataset.

- The `generated` split contains 137,419 high-quality clips synthesized using a fine-tuned [F5-TTS](https://github.com/f5-lab/tts) model, covering 404 equal distribution of synthetic voices.

---

## 📊 Dataset Statistics

| Split       | # Clips   | Duration (hours) |
|-------------|-----------|------------------|
| `generated` | 137,419   | 173.76           |

**Total duration:** ~**173 hours**


## 🛠️ Loading the Dataset

```python
from datasets import load_dataset

dataset = load_dataset("ErikMkrtchyan/Hy-Generated-audio-data-2")
```