obaydata/multilingual-tts-corpus

Multilingual TTS Corpus

A multilingual text-to-speech dataset containing audio recordings with text transcriptions across multiple languages. This is the first batch; more languages will be added over time.

Languages (Batch 1)

Subset	Language	Audio Files	Annotation Format	Source
ru-tts/	Russian	10	JSONL (single file)	俄语TTS（文本标注及语音）基石数据集 #294
ru-speech/	Russian	5	JSONL (single file)	俄语高质量语音音频语料库 #481
vi-speech/	Vietnamese	5	JSONL (single file)	越南语高质量语音音频语料库 #483
ar-tts/	Arabic	10	JSON (per-utterance)	阿拉伯语TTS（文本标注及语音）基石数据集 #210
vi-legal-dialogue/	Vietnamese	13	JSONL (single file)	Vietnamese Legal Consultation Dialogue Dataset

Directory Structure

multilingual-tts-corpus/
├── README.md
├── ru-tts/
│   ├── audio/          # 10 WAV files
│   └── annotations/    # audio.jsonl
├── ru-speech/
│   ├── audio/          # 5 WAV files
│   └── annotations/    # part.jsonl
├── vi-speech/
│   ├── audio/          # 5 WAV files
│   └── annotations/    # part.jsonl
├── ar-tts/
│   ├── audio/          # 10 WAV files
│   └── annotations/    # 10 JSON files (one per utterance)
└── vi-legal-dialogue/
    ├── audio/          # 13 WAV files (multi-turn legal dialogues)
    └── annotations/    # dialogues.jsonl

Audio Format

All audio files share the same format:

• Format: WAV
• Sample Rate: 16,000 Hz
• Bit Depth: 16-bit
• Channels: Mono

Annotation Schema

Each annotation entry contains:

{
  "id": "uuid",
  "audio": {
    "path": "...",
    "format": "WAV",
    "sample_width": "16 bits",
    "sample_rate": "16000 Hz",
    "channels": "1",
    "duration": "N.NN s",
    "snr": "NN dB"
  },
  "labels": {
    "pjwk_cates": {
      "level1": ["category"],
      "level2": ["subcategory"]
    }
  },
  "text": {
    "content": "transcription text..."
  }
}

Key Fields

Field	Description
id	Unique utterance identifier (UUID)
audio.duration	Audio duration in seconds
audio.snr	Signal-to-noise ratio
labels.pjwk_cates	Two-level content category
text.content	Text transcription

vi-legal-dialogue

Vietnamese legal consultation dialogue dataset. 13 multi-turn dialogues covering labor law and contract law topics. Each entry contains the full dialogue transcript (tab-separated turns) paired with audio. Originally from obaydata/vietnamese-legal-dialogue.

• Audio: WAV, 44100 Hz, 16-bit, stereo
• Annotation: dialogues.jsonl — one JSON object per dialogue
• Text format: Multi-turn, each line: <turn>\t<speaker>\t<text>
• Domain: Legal consultation (Labor law, Contract law)
• License: CC-BY-4.0

Adding New Languages

To add a new language batch:

1. Create a new directory: {lang}-tts/ or {lang}-speech/
2. Place WAV files in audio/ and annotations in annotations/
3. Follow the same annotation schema above
4. Update this README table

License

CC-BY-4.0

Citation

@dataset{obaydata_multilingual_tts_2026,
  title={Multilingual TTS Corpus},
  author={ObayData},
  year={2026},
  publisher={Hugging Face},
  url={https://huggingface.co/datasets/obaydata/multilingual-tts-corpus}
}