SafnasKaldeen/Sinhala-Text-speech-Dataset-TTS

Sinhala Text-to-Speech (TTS) Dataset Release

This dataset is the outcome of my Final Year Project at the Department of Computer Science and Engineering, University of Moratuwa. The goal was to build a high-quality, publicly available Sinhala speech corpus to support research and development in speech synthesis and related fields.

About the Author

Safnas Kaldeen
Undergraduate Student – University of Moratuwa, CSE Department
GitHub Repository: https://github.com/SafnasKaldeen/TTSx

Dataset Description

The dataset consists of speech recordings and their corresponding text transcriptions collected from four native Sinhala speakers. Each speaker read a curated set of sentences carefully designed to cover a wide range of phonetic and prosodic features of the Sinhala language. This enables training robust and natural-sounding Text-to-Speech models.

Speaker	Duration (hours)	Description
Isuru	1.22	Clear male voice with neutral tone
Yasindu	2.18	Male voice with natural intonation variations
Harini	2.14	Female voice, expressive and fluent
Dinithi	4.82	Female voice with longer recording duration

Data Collection and Quality

• All recordings were captured using high-quality microphones in a quiet environment to ensure clarity and minimal background noise.
• Transcriptions are manually verified for accuracy and aligned at the sentence level.
• Text data is provided in standard Sinhala Unicode encoding.
• The dataset includes a diverse set of phonemes and common sentence structures to help improve model generalization.

Intended Use Cases

This dataset is designed to support:

• Training Sinhala TTS systems with natural voice synthesis.
• Speaker adaptation and multi-speaker voice modeling.
• Research in prosody, speech segmentation, and phoneme modeling.
• Educational purposes related to speech and language technologies.

How to Access and Cite

The complete dataset, along with preprocessing scripts and training guidelines, is available at the GitHub repository:
https://github.com/SafnasKaldeen/TTSx

If you use this dataset in your research or projects, kindly acknowledge by citing the repository or mentioning Safnas Kaldeen as the dataset author.

---

Thank you for your interest in advancing Sinhala speech technology!