Readable text in a photo
A label on a product, a sign, a page of a document or a screenshot often smears the letters when enlarged. EasyUpscale AI has its own text model, trained from scratch for letters, labels and documents.
Download free for Windows · Buy a license · €14.99
How text is enlarged
- A page of text. Auto mode recognizes a picture that really is text (light paper, thin dark ink in lines) and uses the text model.
- Text in a photo. In an ordinary photo, Auto finds the parts with text (a label, a sign, a page on a table) and enlarges just those with the text model. The rest goes through the photo model and the edges blend softly.
- By hand. In Advanced mode (with a license) you can choose the text model for a whole picture.
Honestly
- The check for what is text is strict on purpose: it finds about three quarters of a label and nothing in photos without text. A missed word is simply enlarged by the photo model.
- The model enlarges, it does not read (it is not OCR). Very blurred text stays unreadable.
- The model is trained only on data that may be used commercially (see About).