OCR demo — single-line (TrOCR) & multi-line (Donut)
Tip: For full screenshots choose a Donut model (Document OCR). For single lines, crop and use TrOCR. Donut models are larger and may take longer to download.
Model:
Xenova/trocr-small-handwritten (small)
Xenova/trocr-base-printed (single-line)
Xenova/donut-base-finetuned-docvqa (DocVQA)
Xenova/donut-base-finetuned-cord-v2 (receipts / forms)
Load model
Model not loaded
Load sample screenshot
Remove image
Run OCR
OCR result (raw JSON):
Decoded text:
Copy Text