---
language:
- en
license: cc-by-4.0
task_categories:
- text-to-speech
tags:
- tts
- speech
- youtube
- closed-captions
pretty_name: hindi
size_categories:
- 100K<n<1M
configs:
- config_name: default
  data_files:
  - split: train
    path: data/train-*.parquet
  default: true
dataset_info:
  config_name: default
  features:
  - name: audio
    dtype: audio
  - name: audio_path
    dtype: string
  - name: text
    dtype: string
  - name: duration
    dtype: float64
  - name: snr
    dtype: float64
  - name: video_id
    dtype: string
  - name: source_url
    dtype: string
  - name: start_time
    dtype: float64
  - name: end_time
    dtype: float64
  - name: speaker_id
    dtype: string
  splits:
  - name: train
    num_examples: 501442
---

# hindi

English TTS dataset collected from YouTube using auto-generated closed captions for transcription.

## Stats

| Metric | Value |
|--------|-------|
| Total rows | 501,442 |
| Total duration | 901.25 hours |
| Sources | 1388 videos |
| Avg segment duration | 6.47s |
| Sample rate | 24kHz mono |
| Format | Sharded parquet with embedded audio |

## Dataset Structure

| Column | Type | Description |
|--------|------|-------------|
| `audio` | Audio | Waveform + sampling rate (24kHz mono WAV) |
| `audio_path` | string | Original audio filename |
| `text` | string | Transcript from YouTube closed captions |
| `duration` | float | Segment duration in seconds |
| `snr` | float | Signal-to-noise ratio (dB) |
| `video_id` | string | YouTube video ID |
| `source_url` | string | Source URL |
| `start_time` | float | Segment start in source audio (seconds) |
| `end_time` | float | Segment end in source audio (seconds) |
| `speaker_id` | string | Speaker identifier |

## Usage

```python
from datasets import load_dataset

ds = load_dataset("somu9/hindi", split="train")

sample = ds[0]
print(sample["text"])
print(sample["audio"])  # {"path": ..., "array": [...], "sampling_rate": 24000}
```

## Collection

1. Download YouTube audio and auto-generated closed captions
2. Parse caption timestamps and merge short cues into 3-8 second segments
3. Quality filter — SNR, clipping, music detection, speaker overlap

## License

CC-BY-4.0

---

*Last updated: 2026-06-06 14:05 UTC*