Read a page
Drop in a photo of a page, a receipt or a handwritten note, or paste one with ⌘V.
Or try one of mine
TeleOCR is a 1.4 billion parameter vision language model trained to parse documents: printed pages, receipts, phone photos of handwriting, tables and formulas.
Pick what to read. Tables come back in OTSL, a compact table language, and I draw them as a real table. Formulas come back as LaTeX, ready to copy.
Big photos are scaled to about 1.25 megapixels first, with the same bicubic filter the model was trained with. It runs the community ONNX build by stefanj0, on desktop GPUs only. The two samples are made for this page.
Runs on your device. Your photos are never uploaded.
Drop in a photo of a page, a receipt or a handwritten note, or paste one with ⌘V.