Arcana: Sri Lanka's First Sinhala Text-to-Speech Model

Dialog Axiata and Rime Labs have launched Arcana, Sri Lanka’s first enterprise-ready Sinhala text-to-speech model. This new model is now live and available for deployment. Rime Labs handled the engineering, including data and training, while Dialog provided market knowledge, Sinhala language expertise, and quality testing. This collaboration focused on meeting the high standards of Sinhala speakers for pronunciation. The model offers advanced control, including tuning for specific terminology and custom brand voices. Benchmarked against global providers, Arcana addresses the historical underservice of Sinhala by mainstream voice AI platforms. Sinhala is spoken by about 16 million people. Dialog has a history of deploying AI services in Sri Lanka, including a trilingual AI platform. Rime Labs' architecture supports both on-premise and cloud deployment. One leader stated that local-language AI capabilities are essential for making digital services more accessible for Sri Lankan users. Another highlighted the challenges of bringing a new language like Sinhala, which is morphologically complex, to production. The goal is to increase adoption of this model among enterprises and developers, supporting the growth of Sinhala voice AI technologies in Sri Lanka. This matters because it makes digital services more accessible and practical for a large number of Sinhala speakers.