Five languages. Equal rigour.

Igbo has the worst existing model performance of the five Nigerian languages we cover and the deepest dialect fragmentation. Owerri, Onitsha, Nsukka, Nnewi, and Ikwuano Igbo are mutually intelligible but acoustically distinct. We collect 380 validated hours across all major dialect regions.

Yoruba is a tonal language — the same phoneme sequence with different tones means different words. Existing commercial models do not preserve tone. We collect 360 validated hours with explicit tone-mark policy in transcription, drawing contributors from Lagos, Ibadan, and across the South-West.

Hausa is the largest indigenous African language by speaker count and the dominant lingua franca of the Sahel. We collect 300 validated hours from Kano, Kaduna, Katsina, Sokoto, and the broader North, with explicit attention to Islamic finance terminology and rural agricultural domains.

Nigerian Pidgin is the true lingua franca of Nigeria — used by every social class, every region. It has no standard orthography. We adopt the Naijá orthographic convention developed at the University of Port Harcourt and apply it consistently across 240 validated hours.

Nigerian English is a distinct variety — not a deviation from British or American English. Existing models perform reasonably on educated Lagos English but fail on Northern Nigerian English, Middle Belt English, and accented variants. We collect 220 validated hours focused on the underrepresented variants and on code-switching where English is the matrix language.