Nzeru Africa on Hugging Face

Welcome to the Hugging Face organization for Nzeru Africa, an open project building AI models for Africa! 🌍🤗

Nzeru means "wisdom" in most African languages. Most AI today works poorly, or not at all, in African languages and contexts. We are changing that by building open models and datasets with African communities, for everyone to use.

Tokiva: speech models for African languages

Tokiva is our series of open speech models for African languages, built from voices recorded with speakers' consent.

• Tokiva ASR: Automatic speech recognition (ASR) that turns spoken African languages into text, for transcription, voice apps and accessibility tools.
• Tokiva Speech datasets: Open recordings of native speakers, ready for training and evaluating speech models.

Tokiva starts with Chichewa and grows to Tumbuka, Yao, Lomwe, Sena, Tonga and more languages across the continent.

Beyond speech

We also build open models for the sectors that matter most across Africa:

• Health: Models that support health workers and help communities get health information in their own languages.
• Education: Models for learning tools, reading support and educational content in African languages.
• Agriculture: Models that give farmers advice on crops, weather and markets in their own languages.
• More sectors: We are adding new models in other areas that affect daily life.

All our models are openly accessible, so researchers, developers and organizations can use them, build on them and improve them.

Get involved

• Contribute your voice: record sentences in your language on nzeruafrica.org.
• Partner with us: researchers, universities, NGOs, governments and companies are welcome.
• Follow this page to get new models and datasets as soon as we release them.

Resources

• Website
• Hugging Face community
• GitHub