MWire Labs Launches Lemka Speech AI for Six NE Languages

MWire Labs, a Shillong-based research and AI deployment lab, has launched *Lemka*, an end-to-end speech AI system combining speech-to-text and text-to-speech capabilities for six languages of Northeast India: Khasi, Garo, Mizo, Meitei (Manipuri), Nagamese, and Kokborok.

The system allows users to speak and be understood in their own language, and receive spoken responses in return, without needing to type, read, or know English or Hindi. Envisioned use cases include government helplines, agricultural advisories for farmers, and voice-driven learning tools for students. Built on over 1,000 hours of field-recorded audio, Lemka is designed for offline and edge deployment in low-connectivity areas, extending the lab's earlier offline voice assistant work for Garo.

Lemka builds on MWire Labs' prior release of *NE-ASR*, an open speech recognition model for eight Northeast Indian languages, and is part of its broader *NE-Stack* initiative aimed at building foundational AI infrastructure for the region's 220-plus indigenous languages. The lab is now pursuing pilot deployments with government and institutional partners.