| --- |
| title: lab.et — The AI Lab of Ethiopia |
| emoji: 🎙️ |
| colorFrom: red |
| colorTo: yellow |
| pinned: true |
| license: cc-by-4.0 |
| --- |
| |
| # 🇪🇹 lab.et — The AI Lab of Ethiopia |
|
|
| <div align="center"> |
|
|
| **We build artificial intelligence for Ethiopian languages.** |
|
|
| *በእኛ ቋንቋ የሚሰራ አስተውሎት።* |
|
|
| [🌐 lab.et](https://lab.et) · [🎙️ voice.et](https://voice.et) · [✨ yakal.et](https://yakal.et) |
|
|
| </div> |
|
|
| --- |
|
|
| ## Who we are |
|
|
| [lab.et](https://lab.et) is an artificial intelligence lab based in **Addis Ababa, Ethiopia**. Amharic is spoken by over 50 million people — yet it remains one of the world's most under-resourced languages in AI. We exist to change that. |
|
|
| We collect our own datasets. We train our own models. We run our own infrastructure. No rented intelligence, no foreign data flows — Ethiopian AI, built in Ethiopia. |
|
|
| ## What we build |
|
|
| ### 🎙️ [voice.et](https://voice.et) — Speech AI |
| Automatic speech recognition and natural text-to-speech for **Amharic, Afaan Oromoo, Tigrinya, Sidamo and Wolaytta** — in production, serving real users through APIs and live voice products. |
|
|
| ### ✨ [yakal.et](https://yakal.et) — The AI Console |
| Chat, images, video, voice and an AI site builder — in Amharic, Afaan Oromoo and English, **prepaid in Birr** via Telebirr. Built for Ethiopian creators, developers and businesses. |
|
|
| ### 🏢 Enterprise & Research |
| Custom speech models tuned to your domain, on-premise and sovereign deployments that keep Ethiopian data inside Ethiopia, and ethically-sourced language corpora for research institutions. |
|
|
| ## 📦 Open releases |
|
|
| | Release | What it is | |
| |---|---| |
| | [azemari](https://huggingface.co/Lab-et/azemari) | 🎙️ **Our flagship TTS** — multilingual Ethiopian voice: Amharic, Afaan Oromoo, Tigrinya · emotion control · voice cloning · complete checkpoint, ready to run | |
| | [omnivoice-amharic](https://huggingface.co/Lab-et/omnivoice-amharic) | Our in-house Amharic text-to-speech system | |
| | [amharic-corpus](https://huggingface.co/datasets/Lab-et/amharic-corpus) | **408.5 hours** of clean synthetic Amharic speech (146,359 clips) — generated entirely by OmniVoice | |
| | [azemari-samples](https://huggingface.co/datasets/Lab-et/azemari-samples) | Listening samples for Azemari — languages, emotions, voice cloning | |
| | *more coming soon* | datasets, models, benchmarks for Ethiopian languages | |
|
|
| ## By the numbers |
|
|
| - 🗣️ **5 languages** supported in production ASR |
| - 📼 **~193k** speech utterances in our training corpora |
| - 🔬 **In-house** model training, evaluation and research pipelines |
| - 🏭 **Production-proven** — live products serving real users since 2025 |
|
|
| ## Work with us |
|
|
| Building for the Ethiopian market? Banks, telcos, government services, NGOs — reach us: |
|
|
| - 🌐 [lab.et](https://lab.et) |
| - ✉️ [admin@lab.et](mailto:admin@lab.et) |
|
|
| We read everything. |
|
|
| --- |
|
|
| <div align="center"> |
|
|
| **lab.et** · Built in Addis Ababa · *በአዲስ አበባ የተሰራ* 🇪🇹 |
|
|
| </div> |
|
|
