Frequently asked questions

Yes. Upload a PDF and each page is analyzed separately. Pages with handwriting are cleaned; pages with no handwriting at all are left completely untouched — including their original vector text where the PDF has it — instead of being needlessly reprocessed.

Yes. JPG, PNG, WEBP, and TIFF photos are supported. Phone photos are automatically rotated using their embedded orientation data. Note that we do not yet correct tilt or perspective distortion in a photo — a heavily skewed photo is processed as-is.

The system is built specifically to avoid this. It only removes pixels inside regions the AI identifies as handwriting, using a tight, locally-adaptive mask rather than blanking a whole area — so printed text, equations, diagrams, and tables outside those regions are left byte-for-byte identical to your original file.

Yes — handwritten answers, calculations, checkmarks, circles, and grading marks are exactly what the detection model looks for. If handwriting is written directly on top of printed text, the area it covers is reconstructed from the surrounding background, since the covered printed content is not otherwise recoverable from a single image.

Yes, scanned documents are a primary use case, including multi-page scanned PDFs and single scanned images.

No. There is no sign-up, login, or account of any kind. Upload a file and download your result.

PDF, JPG/JPEG, PNG, WEBP, and TIFF.

Your document is never uploaded. It is read, edited and cleaned entirely inside your browser, so there is no server copy, no database and nothing that could be used to train a model. Closing the tab discards it.

You brush over the handwriting you want removed. Inside that mark, ink pixels are identified against the local paper background, printed content such as question text and diagram lines is detected and protected, and the remaining ink is reconstructed from the surrounding untouched pixels. Everything outside your mark is never modified.

Yes, as a distinct step: native PDF 'Ink' annotations (freehand strokes added in a PDF editor) are removed directly from the PDF's own annotation data before any image processing happens, which is lossless. This is different from handwriting physically scanned or photographed onto a page, which is removed via the AI + inpainting pipeline described above.