Skip to main content
Region-first speech and text models. Open NE-ASR (and other weights) on Hugging Face. LEMKA, TTS voices, and the HTTP API ship on a licensed box. There is no public multi-tenant cloud.

Speech-to-text

Multi ASR for Khasi, Garo, Kokborok, Mizo, Nagamese, Meitei.

Text-to-speech

One voice per language. Licensed with the image.

Text

Khasi, Garo, Mizo, Meitei. Translate and embeddings. Limited.

On-prem

Dockerised models. Data stays on the box.

HTTP API

Live demo routes now. Versioned /v1 on request.

Request access

Email connect@mwirelabs.com

​
Pick a path

Run weights

Hugging Face checkpoints for local inference.

Run the API

GET /health and POST /speak on a licensed box.
The current HTTP demo is Garo-first (POST /speak). Multi-language /v1 routes ship with a licensed on-prem image.