Discover and experiment with our state-of-the-art voice models

Every model here is built speech-first, for languages the industry has passed over. Each one is trained on recordings given with consent by the communities that speak them, and shaped to work on the channels people already reach for.

Watercolour illustration of two speech bubbles, one marked HA and one marked FR, joined by flowing sound waves
Speech translation

Namu-Interpret

Hausa and French, carried both ways in speech. One side talks, the other hears it in their own language, and the conversation keeps its pace.

Watercolour illustration of a sound wave resolving into four written lines of text
Speech recognition

Namu-Transcribe

Turn Hausa speech into accurate text, across accents, dialects and the noisy rooms people record in.

Watercolour illustration of a megaphone with sound radiating outward from it
Speech synthesis

Namu-Voice

Natural, expressive Hausa speech from written text, shaped to be understood the first time it is heard.

Watercolour illustration of two overlapping speech bubbles with a sound wave running between them
Voice agent

Namu-Agent

An end-to-end conversation in Hausa. It listens, works out what you need, and answers out loud.