Run it
Drop / choose your own image
–text regions
backend –latency –
← All models · Scene text detection
Basics · object-detection (text) · WASM
The plain capability: give the model an image, get a box around every region it thinks is text, and a count. No reading — just where.
–text regions
Each indigo polygon is one connected region of high text-probability. The model separates a sign's lines, a receipt's items, even a headline from its subtitle. It never tells you what the text says — that's the next step. Detection is the front door to every OCR pipeline.