A curated corpus for building and fine-tuning language models that can converse in, translate, and help preserve two low-resource Gabonese languages: Fang and Punu (Yipunu), bridged through French and English. These languages have very little digital text. This dataset is a preservation artifact: a cleaned corpus assembled from dictionaries, grammars, example sentences, and religious texts, plus ready-to-train instruction/LM splits.… See the full description on the dataset page: https://huggingface.co/datasets/vmombo/gabon-fang-punu.