ParsVoice: A Large-Scale Multi-Speaker Persian Speech Corpus for Text-to-Speech Synthesis

2025-10-12

AI Summary

Despite over 100 million native speakers, Persian suffers from a severe scarcity of high-quality, multi-speaker speech corpora, hindering the development of text-to-speech (TTS) and other speech technologies. To address this low-resource challenge, we propose a scalable methodology for constructing a large-scale Persian TTS corpus: an automated pipeline integrating BERT-based sentence completion detection, binary-search-based boundary optimization, and a multidimensional quality assessment framework—enabling precise audio-text alignment and robust data filtering. Leveraging 2,000 Persian audiobooks, we curate 3,526 hours of cleaned speech and release a high-fidelity subset of 1,804 hours, covering 470+ speakers. To our knowledge, this is the largest and highest-quality open-source Persian TTS corpus to date, substantially alleviating the data bottleneck for Persian speech synthesis and related tasks.

Technology Category

• Natural Language Processing: Speech
• Machine Learning: Large Multimodal Models (LMMs)
• Data Mining & Knowledge Management: Conversational Systems for Recommendation & Retrieval

Application Category

• Search and Retrieval-Augmented AI: Multilingual and cross-lingual Web search
• Web Mining and Content Analysis: Mining multimedia, multimodal, multilingual, cross-lingual Web data
• Economics, Online Markets and Human Computation: Data quality aspects of human-annotated datasets

Abstract

Persian Language, despite being spoken by over 100 million people worldwide, remains severely underrepresented in high-quality speech corpora, particularly for text-to-speech (TTS) synthesis applications. Existing Persian speech datasets are typically smaller than their English counterparts, which creates a key limitation for developing Persian speech technologies. We address this gap by introducing ParsVoice, the largest Persian speech corpus designed specifically for TTS applications. We created an automated pipeline that transforms raw audiobook content into TTS-ready data, incorporating components such as a BERT-based sentence completion detector, a binary search boundary optimization method for precise audio-text alignment, and multi-dimensional quality assessment frameworks tailored to Persian. The pipeline processes 2,000 audiobooks, yielding 3,526 hours of clean speech, which was further filtered into a 1,804-hour high-quality subset suitable for TTS, featuring more than 470 speakers. ParsVoice is the largest high-quality Persian speech dataset, offering speaker diversity and audio quality comparable to major English corpora. The complete dataset has been made publicly available to accelerate the development of Persian speech technologies and to serve as a template for other low-resource languages. The ParsVoice dataset is publicly available at ParsVoice (https://huggingface.co/datasets/MohammadJRanjbar/ParsVoice).

Problem

• Addressing the scarcity of high-quality Persian speech corpora for TTS
• Developing an automated pipeline to process raw audiobooks into TTS data
• Creating the largest multi-speaker Persian dataset to enable speech technologies

Innovation

• Automated pipeline transforms audiobooks into TTS data
• BERT-based sentence completion detector for accurate segmentation
• Binary search method optimizes audio-text alignment precision