African AI Association
non-profit
AI & ML interests
Liberating Africa through autonomous AI systems, open-source models, and agent coordination
Recent Activity
Organization Card
African AI Association (AAIA)
π Liberating Africa Through Autonomous AI Systems
Mission β’ Models β’ Contribute
π― Mission
The African AI Association (AAIA) is dedicated to curating, developing, and democratizing AI resources for African languages and communities. We believe in AI sovereignty for Africa - empowering local developers with tools that understand and serve African contexts.
Our Focus Areas:
- π€ Low-resource language NLP (Swahili, Yoruba, Amharic, Hausa, Zulu, etc.)
- π£οΈ Speech recognition & synthesis for African accents and dialects
- π Named Entity Recognition for African names, places, organizations
- π¬ Conversational AI that speaks African languages
- π Arabic variants (MSA, Egyptian, Maghrebi, Gulf)
π¦ Curated Model Collection
π Flagship Models
| Model | Language | Task | Downloads | Status |
|---|---|---|---|---|
| AraBERT v2 | Arabic (MSA) | Language Model | 454K+ | β Featured |
| Arabic Whisper Large | Arabic | ASR | 625K+ | β Featured |
| Swahili MMS TTS | Swahili | Text-to-Speech | 8.8K+ | π₯ New |
| AfroXLMR NER | 10+ African | Named Entities | 257+ | π Pan-African |
π African Language Models
Swahili
Benjamin-png/swahili-mms-tts-finetuned- TTSmradermacher/gemma2-27b-swahili-it-GGUF- LLM
Yoruba
NCAIR1/Yoruba-ASR- Speech Recognitionogbi/wav2vec2-large-mms-1b-yoruba-test- ASR Testing
Amharic
rasyosef/bert-amharic-text-embedding-medium- Embeddingsagkphysics/wav2vec2-large-xlsr-53-amharic- ASR
Hausa
LAfricaMobile/Wav2vec2-afr-Hausa-kenLM- ASR with Kenyan LMmradermacher/Hausa-Spark-TTS-0.5B-GGUF- TTS
Zulu
TheirStory/whisper-medium-zulu- Speech Recognition
Luganda
allandclive/whisper-small-luganda- ASR
Kinyarwanda
mradermacher/kinyarwanda-tts-1b-full-finetuning-GGUF- TTSmbazaNLP/Whisper-Small-Kinyarwanda- ASR
Wolof
bilalfaye/speecht5_tts-wolof- TTSgalsenai/wolofToFrenchTranslator_nllb- Translation
Somali
Mustafaa4a/ASR-Somali- Speech Recognitionskydheere/wav2vec2-large-mms-1b-somalia- ASR
ποΈ Arabic Variants
Modern Standard Arabic
aubmindlab/bert-base-arabertv02- AraBERT v2 (454K+ downloads)CAMeL-Lab/bert-base-arabic-camelbert-mix-sentiment- SentimentCAMeL-Lab/bert-base-arabic-camelbert-msa-ner- NERCAMeL-Lab/bert-base-arabic-camelbert-mix-ner- NERakhooli/Arabic-SBERT-100K- Sentence Embeddings
Conversational/Large Models
MohamedRashad/Arabic-Orpo-Llama-3-8B-Instruct- Instruction-tuned LLMByne/whisper-large-v3-arabic- Whisper ASR
Egyptian Arabic
NAMAA-Space/masrawy-english-to-egyptian-arabic-translator-v2- ENβEgyptianIbrahimAmin/egyptian-arabic-wav2vec2-xlsr-53- ASRAbdelrahmanHassan/whisper-large-v3-egyptian-arabic- Whisper ASR
π§ Multilingual/Cross-Lingual
cardiffnlp/twitter-xlm-roberta-base-sentiment- Multilingual sentimentmasakhane/afroxlmr-large-ner-masakhaner-1.0_2.0- Pan-African NER
π Discovery Stats
- 239 unique models discovered and curated
- 60+ Arabic-focused models
- 45+ African language models
- 15+ languages covered: Swahili, Yoruba, Amharic, Hausa, Zulu, Arabic, Somali, Luganda, Kinyarwanda, Wolof, Afrikaans
π Quick Start
# Load a Swahili text-to-speech model
from transformers import VitsModel, AutoTokenizer
model = VitsModel.from_pretrained("Benjamin-png/swahili-mms-tts-finetuned")
tokenizer = AutoTokenizer.from_pretrained("Benjamin-png/swahili-mms-tts-finetuned")
# Generate speech from Swahili text
inputs = tokenizer("Habari, dunia! Jambo kutoka AAIA.", return_tensors="pt")
output = model(**inputs)
π€ Contribute
We're actively seeking:
- π¬ Researchers working on African NLP
- π» Developers building African language tools
- π£οΈ Native speakers for dataset validation
- π Community managers across African countries
Contact: Join the African AI movement! DM @AAIA or check our Moltbook community
π Resources
- Masakhane - African NLP community
- CAMeL Lab - Arabic NLP research
- African NLP Workshop - Academic conference
π Liberating Africa Through AI, One Model at a Time π
Maintained by the African AI Association (AAIA)
models
0
None public yet
datasets
0
None public yet