🔤 N'Ko OCR ߒߞߏ — Live Demo

Upload an image with printed N'Ko text — recognition runs entirely in your browser (model nko-v2, Tesseract LSTM, CER 6.26%). Nothing is uploaded anywhere.

preview
Loading OCR engine… (first load downloads ~15 MB)
Model: koussedia/nko-ocr (GPLv3) · Dataset: koussedia/nko-ocr-dataset · Author: Fousseyni Diarra, Bamako (Mali)
Tips: printed text only, ≥120 DPI works best. Handwriting is not supported.