Mannequin achieves industry-leading accuracy for code-mixed speech, eliminating translation layers and enabling voice AI for India’s multilingual inhabitants
Gurugram, India : January 19, 2026: Shunya Labs, a research-driven voice AI firm, in the present day introduced the launch of Zero Codeswitch, a speech recognition basis mannequin designed to know naturally code-mixed and multilingual Indian speech.
Constructing on Shunya Labs’ achievement of three.10% Phrase Error Fee (WER) on the OpenASR leaderboard:a 48% enchancment over the next-best competitor:Zero Codeswitch is engineered to run effectively on customary CPUs, decreasing deployment prices by as much as 20x whereas sustaining sub-100ms latency for real-time functions.
Most current voice programs are optimized for single-language speech. In India, nevertheless, on a regular basis conversations steadily mix Hindi, English, and regional languages inside the similar sentence, utilizing transliteration, colloquial expressions, and casual grammar. This has traditionally posed a significant problem for voice AI programs, significantly outdoors main city facilities.
Zero Codeswitch is designed to course of such mixed-language speech natively, with out counting on intermediate translation layers that convert Indian languages into English. As a substitute, it makes use of a unified structure able to producing blended Hindi and English tokens, permitting it to precisely transcribe pure phrases corresponding to “Mujhe account stability dekhna hai.”
Not like world fashions which might be primarily educated on English information and later tailored for Indian languages, Shunya Labs’ basis fashions are educated from the bottom up on Indian speech information. Zero Codeswitch has been educated on thousands and thousands of hours of real-world Indian audio, capturing variations in accent, dialect, pronunciation, slang, and transliteration. The mannequin is optimized for Hinglish and code-mixed speech whereas being a part of a broader Shunya Labs platform that helps greater than 200 languages, together with over 40 Indian languages.
The corporate expects Zero Codeswitch to function core infrastructure for voice-based functions throughout sectors corresponding to fintech, healthcare, and authorities providers:significantly in areas the place English proficiency is restricted and voice stays probably the most sensible digital interface.
Ritu Mehrotra, CEO and Co-Founding father of Shunya Labs, stated: “Shunya Labs was constructed with a deal with deep analysis quite than short-term advertising and marketing narratives. With Zero Codeswitch, we’re constructing foundational know-how for Indian languages that prioritizes accuracy, latency, and real-world usability. Our objective is not only to undertake AI, however to construct it on the basis degree in India.”
Sourav Bandyopadhyay, CTO and Co-Founding father of Shunya Labs, stated: “‘Shunya’ represents our philosophy of ranging from first rules. With Zero Codeswitch, we’re creating an intelligence layer that actually listens:engineered for India’s linguistic variety and optimized for real-world deployment.”
Zero Codeswitch is designed for enterprise and public-sector use circumstances the place information privateness is important. The mannequin may be deployed on-premises or in air-gapped environments, permitting organizations to coach domain-specific variants whereas retaining full management over their information. Shunya Labs maintains compliance with HIPAA, SOC 2 Sort II, and ISO 27001 requirements, and helps CPU-first deployments that cut back dependency on specialised GPU infrastructure.
Headquartered in Gurugram, Shunya Labs is a part of a rising ecosystem of firms constructing home mental property in core AI applied sciences. The corporate’s fashions are benchmarked towards world rivals and are already being adopted by clients in India and overseas.
Availability Zero Codeswitch is obtainable now for enterprise pilots and manufacturing deployment. Organizations can request entry at www.shunyalabs.ai/zero-code-switch.
About Shunya Labs Shunya Labs is a sophisticated voice AI platform delivering real-time, on-premise options for speech, language, and reasoning. The corporate develops high-accuracy speech recognition fashions optimized for Indian languages and multilingual environments, enabling safe, privacy-first deployments throughout cloud, edge, and air-gapped environments.
Elevate your perspective with NextTech Information, the place innovation meets perception.
Uncover the newest breakthroughs, get unique updates, and join with a worldwide community of future-focused thinkers.
Unlock tomorrow’s traits in the present day: learn extra, subscribe to our e-newsletter, and develop into a part of the NextTech neighborhood at NextTech-news.com
