Pixel mosaic in Nzinga colours, one bright cell at its core
Docs
/
Start

Nzinga Docs

Plain-language guides for everyone who works with Nzinga. Pick the page written for you.

8 guides
Plain language
01

What Nzinga is

Nzinga is language infrastructure for African AI. It has three parts.

nzinga
in progress
Nzinga Tokenizer: fewer tokens for African languages
measured
Nzinga App: invited members record their languages
demo
Nzinga Network: a consented, paid voice library
pilot planned
Sources: Nzinga tokenizer README, Nzinga README

The tokenizer cuts the extra tokens African languages pay. The app lets invited members record voice notes and choose who can use each one. The network turns consented recordings into datasets and pays members when recordings are accepted.

Note

The technical reference lives in the developer documentation. These pages explain what Nzinga does and how to take part, for readers who are not engineers.

03

Where things stand

Every guide keeps three things apart: what is measured, what is built and what is planned.

status
today
Tokenizer token counts on two held-out test sets
measured
Consent rules, invites and the earnings ledger
built
The phone app, with a demo mode
built
Audio upload, payouts and datasets
planned
Model accuracy with the Nzinga tokenizer
not yet measured
Sources: tokenizer README, Nzinga README, proof-of-concept plan
Note

Where a guide says planned or to be decided, the owner has not made that call yet.

04

Get in touch

Nzinga is private and invite-only. To ask a question, propose a pilot or request access, use the contact page or the project on GitHub.