README.md

---
license: cc-by-4.0
language:
- hi
- en
tags:
- audio
- speech
- tts
- asr
- hinglish
- hindi
- indian-english
- text-to-speech
- automatic-speech-recognition
- code-switching
pretty_name: Hinglish Concatenated Audio Dataset
size_categories:
- 100K<n<1M
task_categories:
- automatic-speech-recognition
- text-to-speech
annotations_creators:
- expert-generated
- crowdsourced
language_creators:
- crowdsourced
- expert-generated
multilinguality:
- multilingual
---

# Hinglish Concatenated Audio Dataset

A large-scale, cleaned and annotated speech dataset covering **Hindi**, **Hinglish** (Hindi–English code-switching), and **Indian English** — compiled from 14 public corpora and original custom recordings, unified into a single Parquet dataset with consistent schema.

## At a Glance

| Stat | Value |
|------|-------|
| **Total clips** | 815,171 |
| **Total Estimated Hours** | 2,264+ |
| **Unique speakers** | 6,304 |
| **Raw audio size** | ~243 GB |
| **Languages** | Hindi (`hi`), Hinglish (`hi-en`), Indian English (`en-IN`) |
| **Format** | Parquet with embedded audio (HF dataset viewer compatible) |
| **Tasks** | ASR, TTS fine-tuning, voice cloning, speech research |

---

## Dataset Schema

| Column | Type | Description |
|--------|------|-------------|
| `source` | `string` | Speaker or source identifier |
| `text` | `string` | Cleaned transcript (may include `<hi-en>` language tags) |
| `audio` | `Audio` | WAV audio embedded in Parquet with native sample rate |
| `quality` | `string` | Sample rate (Hz) or MOS quality score where available |
| `duration` | `string` | Clip duration in seconds |

---

## Source Breakdown

| Source Dataset | Clips | Description | License |
|----------------|------:|-------------|---------|
| [NPTEL Hindi Spoken Tutorial](https://huggingface.co/datasets/ai4bharat/Spoken-Tutorial) | 521,028 | Educational lectures in Hindi — AI4Bharat / NPTEL | CC BY 4.0 |
| [AI4Bharat Kathbath](https://huggingface.co/datasets/ai4bharat/Kathbath) | 94,903 | Multi-speaker Hindi speech benchmark | CC BY 4.0 |
| [AI4Bharat IndicTTS](https://huggingface.co/datasets/SPRINGLab/IndicTTS-Hindi) | 36,613 | Studio-quality Hindi TTS corpus | CC BY 4.0 |
| [Hinglish — ujs](https://huggingface.co/datasets/ujs/hinglish-compressed) | 25,378 | Code-switched Hindi–English speech | See source |
| [Mozilla Common Voice 17 (Hindi)](https://huggingface.co/datasets/mozilla-foundation/common_voice_17_0) | 24,643 | Community-contributed Hindi speech | CC0 1.0 |
| [AI4Bharat Mann Ki Baat](https://huggingface.co/datasets/ai4bharat/Mann-ki-Baat) | 22,483 | Broadcast Hindi speech, En-Indic aligned | CC BY 4.0 |
| Mann Ki Baat (English) | 22,477 | English side of Mann Ki Baat parallel corpus | CC BY 4.0 |
| [Hindi Female Single Speaker HQ — Shekharmeena](https://huggingface.co/datasets/Shekharmeena/Hindi-Female-Single-Speaker-High-Quality-Speech) | 22,058 | High-quality single female speaker Hindi | See source |
| [Orpheus TTS Indian English](https://huggingface.co/datasets/ar17to/orpheus_tts_english_indian_multispeaker) | 18,238 | Multi-speaker Indian English TTS | See source |
| [Indic TTS Hindi — SPRINGLab](https://huggingface.co/datasets/SPRINGLab/IndicTTS-Hindi) | 11,825 | Multi-speaker Hindi TTS | CC BY 4.0 |
| [Indian English — krishan23](https://huggingface.co/datasets/krishan23/indian_english) | 6,765 | Indian-accented English speech | See source |
| [Hinglish Test TTS — Shekharmeena](https://huggingface.co/datasets/Shekharmeena/hinglish-test-tts) | 3,136 | Custom Hinglish TTS recordings | See source |
| Custom female TTS recordings | 2,843 | Original studio recordings | CC BY 4.0 |
| [Orpheus TTS Shaurya — prashantarya](https://huggingface.co/datasets/prashantarya/orpheus-tts-Shaurya-dataset) | 1,419 | Male Hindi/English TTS voice | See source |
| [Anika Voice — Shekharmeena](https://huggingface.co/datasets/Shekharmeena/Anika-voice) | ~5 | Custom female Hindi/Hinglish voice | See source |
| **Total** | **815,171** | | |

---

## Curation & Processing

- All clips re-segmented to remove silence and cross-talk
- Transcripts normalised to Unicode NFC; language tags added (`<hi-en>` for code-switched utterances)
- Duplicate and near-duplicate clips removed
- Consistent CSV schema applied across all sources before concatenation
- Audio stored as WAV, original sample rates preserved

---

## License

This dataset is released under **[Creative Commons Attribution 4.0 International (CC BY 4.0)](https://creativecommons.org/licenses/by/4.0/)**.

**You are free to:**
- Use for research, commercial, or personal projects
- Share and redistribute in any medium
- Build upon and adapt the data

**Under these conditions:**
- **Attribution** — Credit this dataset and link to the original source datasets listed above
- **No misrepresentation** — Do not use this data to generate non-consensual synthetic voices of real individuals, or for fraud, harassment, or disinformation

> ⚠️ **Upstream licenses:** Each constituent dataset retains its original license. Users are responsible for complying with the terms of the sources they use. Mozilla Common Voice is CC0; all AI4Bharat and SPRINGLab sources are CC BY 4.0. Sources marked "See source" should be checked individually.

---

## Citation

If you use this dataset in your work, please cite it and the relevant upstream sources:

```bibtex
@dataset{agarwal2026hinglish,
  author    = {Agarwal, Ayushi},
  title     = {Hinglish Concatenated Audio Dataset},
  year      = {2026},
  publisher = {Hugging Face},
  url       = {https://huggingface.co/datasets/agarwalayushi/hinglish},
  note      = {Aggregated from Mozilla Common Voice, AI4Bharat (Kathbath, Mann Ki Baat,
               Spoken Tutorial, IndicTTS), SPRINGLab IndicTTS, and custom recordings}
}
```

---

## Contact

To report transcription errors, audio quality issues, or attribution concerns, please open a [Discussion](https://huggingface.co/datasets/agarwalayushi/hinglish/discussions) on this page.