README.md

---
dataset_info:
  features:
  - name: id
    dtype: string
  - name: audio
    dtype: audio
---

# Audio Dataset

Multi-domain Hindi speech dataset combining human-recorded STT, synthetic TTS, and mixed-source audio clips. All audio is stored as raw bytes and plays directly in the Dataset Viewer.

## Dataset Overview

| HF Subfolder | Description | Files | Size | S3 Source |
| --- | --- | ---: | ---: | --- |
| `stt_hi_dataset_human/ (INCOMPLETE)` | Human-recorded Hindi STT clips | 965,201 | 188.3 GB | `/stt_hi_dataset_human` |
| `tts_hi_dataset_synthetic/` | Synthetic Hindi TTS clips | 39,979 | 7.6 GB | `/tts_hi_dataset_synthetic` |
| `human-and-synthetic/` | Mixed root-level audio (human + synthetic) | 163,182 | 344.4 GB | `/` (root) |

**Total: ~1,168,362 clips · ~540 GB**

## Schema

| Column | Type | Description |
| --- | --- | --- |
| `id` | `string` | Relative file path within the subfolder (no extension) |
| `audio` | `audio` | Raw audio bytes — plays directly in the Dataset Viewer |

---

## Sample Records

### `human-and-synthetic` — audio_dataset_human

```json
{
  "_id": { "$oid": "68ac6a961f0582b904f5edcc" },
  "language": "Bengali",
  "audio_raw_duration": 0.841,
  "scenario": "Read",
  "task_name": "Keywords Spotting",
  "gender": "Female",
  "age_group": "60+",
  "job_type": "Unemployed",
  "qualification": "No Schooling",
  "area": "Urban",
  "district": "Paschim Bardhaman",
  "state": "West Bengal",
  "occupation": "Housewife",
  "verbatim": [{ "end": 0.894, "text": "সুখী", "start": 0.254, "speaker_id": 0 }],
  "normalized": [{ "end": 0.894, "text": "সুখী", "start": 0.254, "speaker_id": 0 }],
  "audio_segmented_duration": 0.64,
  "speaker_id": "S4257764000369908",
  "description": "সুখী",
  "url": "1407374883597003.wav"
}
```

### `human-and-synthetic` — audio_dataset_synthetic

```json
{
  "_id": { "$oid": "68ac68a58f33dfd63eabaeae" },
  "metadata": {
    "language": "Hindi",
    "gender": "Female",
    "sample_rate": 24000,
    "duration": 10.356,
    "text": "Ek classmate se notes lene ke liye call kiya, aur woh 20 minutes tak random baatein share karta raha. Socha, chalo chhodo, at least notes mil gaye na!",
    "audio_info": {
      "word_count": 28,
      "character_count": 150,
      "sentence_count": 2,
      "average_words_per_second": 2.703746620316725
    },
    "dataset_info": {
      "category": "synthetic_speech",
      "tags": ["synthetic", "tts"]
    },
    "url": "sangrah_a996d341f3e04aa88959bd0b4f576a00f7b72e5c79ea.wav"
  },
  "segments": [
    { "text": "Ek classmate se notes lene ke liye call kiya", "start": 0, "end": 2.426, "speaker_id": 0 },
    { "text": "aur woh 20 minutes tak random baatein share karta raha", "start": 2.728, "end": 5.689, "speaker_id": 0 },
    { "text": "Socha", "start": 6.084, "end": 6.583, "speaker_id": 0 },
    { "text": "chalo chhodo", "start": 6.931, "end": 7.744, "speaker_id": 0 },
    { "text": "at least notes mil gaye na!", "start": 8.069, "end": 10.356, "speaker_id": 0 }
  ]
}
```

### `tts_hi_dataset_synthetic`

```json
{
  "_id": { "$oid": "6975056d8d31caaca917995e" },
  "id": "6973a0bf18a0110414658892",
  "audio": "tts_hi_dataset_synthetic/6973a0bf18a0110414658892.wav",
  "duration": 6.48,
  "language": "hi",
  "metadata": {
    "speaker_id": "S0001",
    "gender": "Female",
    "age_group": "20-28",
    "scenario": "Conversation"
  },
  "normalized_text": "आज का weather बहुत unpredictable है [sighs] सुबह तो तेज़ धूप थी और अब sudden change हो गया,",
  "raw_text": "आज का weather बहुत unpredictable है... [sighs] सुबह तो तेज़ धूप थी और अब sudden change हो गया,"
}
```

### `stt_hi_dataset_human`

```json
{
  "_id": "00023bad801142b4b165962d366086cf",
  "audio": "stt_hi_dataset_human/00023bad801142b4b165962d366086cf.wav",
  "duration": 5.16,
  "language": "hi",
  "metadata": {},
  "normalized_text": "मानो पहले से कहीं ज्यादा electricity बढ़ जाती है उसकी पूरी body में blue color की.",
  "raw_text": "मानो पहले से कहीं ज्यादा इलेक्ट्रिसिटी बढ़ जाती है उसकी पूरी बॉडी में ब्लू कलर की"
}
```