Oriserve Unveils Tarang Voice Engine at India AI Impact Summit
The company said Tarang outperforms domestic and global benchmarks in speed, accuracy and dialect comprehension.
Oriserve introduced Tarang, a next-generation Speech-to-Text (STT) and Text-to-Speech (TTS) engine built specifically for India’s diverse linguistic landscape at the India AI Impact Summit.
The company said Tarang outperforms domestic and global benchmarks in speed, accuracy and dialect comprehension.
In recent audits, Tarang recorded a Word Error Rate (WER) of 15% and a Character Error Rate (CER) of 7%, surpassing prevailing regional standards and outperforming many global general-purpose APIs in dialect-heavy environments.
The engine also delivers a Time to First Token (TTFT) of 100 milliseconds, significantly reducing latency and enabling near real-time, natural conversations across 20 of India’s 22 scheduled languages.
To demonstrate real-world resilience, Oriserve showcased a live demo from a drug de-addiction helpline, where Tarang processed slurred speech and background noise to provide accurate guidance.
"Most AI breaks when the user stops speaking 'textbook' Hindi or English. But Bharat doesn't speak in textbooks," said Anurag Jain, co-founder & CEO of Oriserve. "By mastering the dialects and the emotional chaos of real-world interactions, we are moving beyond 'chatbots' to create a 'Zero-UI' future. Our mission is to ensure that for the next 500 million users, voice is the only interface they will ever need to access the digital economy."
Tarang is now available for enterprise integration across helplines, commerce platforms and financial services applications. The company, which has already served enterprise giants like Vodafone Idea and GroMo, is now opening access for developers and enterprises to experience these benchmarks firsthand.
