Addis AI

Voice-first AI infrastructure for African languages.

Addis AI builds speech, voice, and language models for African languages.

This Hugging Face organization is the home for our open models, datasets, and research releases. Our work focuses on building language technology for languages that have historically had limited representation in modern AI systems.

Our current open-source work includes Amharic and Afaan Oromo, with additional African languages being developed.

Open Models

We develop and release models across speech and language, including:

• Large language models
• Automatic speech recognition models
• Text-to-speech and voice models
• Language-specific and multilingual models
• Continued-pretraining and instruction-tuned models

Official Addis AI model releases will be available through this organization and our curated model collections.

Browse Addis AI models

Open Datasets

We release datasets that support research and model development for African languages.

Our open datasets include:

• Speech datasets
• Text corpora
• Instruction-tuning datasets
• Translated datasets
• Language-specific training and evaluation data

We aim to document dataset provenance, creation methodology, intended use, limitations, and licensing directly in each dataset card.

Browse Addis AI datasets

Research Focus

Our work currently spans:

• Automatic Speech Recognition (ASR)
• Text-to-Speech (TTS)
• Large Language Models (LLMs)
• Machine Translation
• Multilingual NLP
• Multimodal AI
• Low-resource language modeling
• African language datasets
• Speech and language evaluation

We are particularly interested in the infrastructure required to make high-quality AI models available for languages with limited training data, evaluation resources, and existing tooling.

Languages

Our current work includes:

• Amharic
• Afaan Oromo

We are also developing and researching additional African languages.

Individual model and dataset cards contain the language coverage specific to each release.

Using Our Open Releases

Each model and dataset repository contains its own documentation covering usage, licensing, limitations, and technical details.

For production APIs, hosted models, and developer tools, use the Addis AI platform rather than assuming that every Hugging Face repository represents the latest production model.

Build with Addis AI

Platform
https://addisassistant.com

Open Source
https://addisassistant.com/open-source

Developer Documentation
https://docs.addisassistant.com

Enterprise Deployments
https://addisai.ch

About Addis AI

Addis AI is building voice-first AI infrastructure for African languages.

Our infrastructure includes models for speech-to-text, text-to-speech, large language models, translation, multimodal reasoning, and real-time voice, available through APIs and official SDKs.

Alongside our production infrastructure, we release selected models and datasets openly to support developers, researchers, and the broader African-language AI ecosystem.

Model and dataset licenses vary by repository. Please review the license and usage information provided on each individual repository before use.