← All models · TrOCR

use case · multi-model

Multi-model: read → translate

One model can't do this alone. First TrOCR reads the text out of the image; then M2M100 translates that text into the language you pick. Pixels → characters → another language, chained end-to-end, all on your device.

Stage 1 — OCR (TrOCR small printed, ~185 MB)

Stage 2 — translation (M2M100 418M, ~470 MB)

Drop an image, or click to choose
A printed line: The quick brown fox jumps A sign reading Platform 9 A receipt line reading Grand Total $42.99

Drag to box a line of text.

1 · Read (English)

2 · Translation

Why chain them? OCR frees text from an image; translation makes it readable to anyone. Together they're a private, offline "point-and-understand" tool — no text, and no image, ever leaves the device. (TrOCR small-printed is English-trained, so this reads English and translates outward.)

← Wild · Back to TrOCR