African-Whisper vs AfriHuBERT

An open-source framework (PyPI: africanwhisper) for fine-tuning OpenAI's Whisper on multilingual African-language audio datasets such as Common Voice and FLEURS, with optimized inference, diarization and deployment. Created by Kevin Kibe.

AfriHuBERT is a compact self-supervised speech representation model based on mHuBERT-147, continually pretrained via multilingual adaptive finetuning on over 10,000 hours of speech spanning more than 1,200 African languages and varieties. It improves spoken language identification and ASR over its base model and acts as an encoder for downstream African speech tasks. Its training data was aggregated from sources including BibleTTS, Kallaama, NaijaVoices and NCHLT.