grushaaaaa/tts-indian

---
license: cc-by-4.0
language:
- bn
- ks
- gu
- kn
- ta
- te
tags:
- tts
- speech
- indian-languages
- audio
pretty_name: TTS Indian Languages
---

# TTS Indian Languages Dataset

Speech dataset for Text-to-Speech covering 6 Indian languages, collected and processed from YouTube.

## Languages & Speakers

| Speaker | Language | Gender |
|---------|----------|--------|
| monihara_bengali | Bengali | Male |
| munir_kashmiri | Kashmiri | Male |
| nandini_gujarati | Gujarati | Female |
| sansri_kannada | Kannada | Female |
| tamil_pokkisham | Tamil | Male |
| teluguM | Telugu | Male |

## Pipeline

Audio was collected and processed through these stages:

1. **YouTube Download** — yt-dlp from curated speaker playlists
2. **Vocal Separation** — Demucs (htdemucs_ft) to remove background music/noise
3. **Audio Enhancement** — Resemble-Enhance + DeepFilterNet for clean speech
4. **Diarization** — Silero VAD + WavLM clustering to segment speaker turns
5. **Chunking** — 3–15 second clips
6. **Quality Filtering** — DNSMOS scoring to keep only clean audio
7. **Transcription** — ASR transcription saved alongside each clip

## Format

Each row contains:
- `audio`: 16kHz mono WAV
- `transcription`: verbatim text
- `speaker`: speaker ID
- `language`: language name
- `gender`: male/female