Short answer
Add the page photo to Fix Documents (BETA). Check that all four page corners are inside the crop.
Compare Original and Cleaned, then save a new PDF. Run OCR afterward only if you need searchable or selectable text.
When to use Fix Documents
Use this workflow for photographed books, forms and paper documents when the page is not square to the camera, bends near an edge or spine, or has distracting shadows and uneven light. It corrects the page image before you make it searchable or send it to someone else.
Technical noteConfirmed behavior and an important limit.
- What we confirmed
- Document processing stays on the PC. The correction model and OCR language packs are optional components downloaded only when requested and verified before use.
- Important limit
- Fix Documents is a BETA feature. It cannot restore detail hidden by glare, motion blur, missing page edges or an incomplete photo, so every result needs a visual check.
What each correction does
Rotation turns the whole page upright. Perspective correction squares a page photographed at an angle. Page flattening reduces curves in a photographed book or bent sheet. Shadow and lighting correction evens the visible page. OCR is a later, separate step that adds a draft text layer; it does not correct the photograph.
What you need before you start
- An untouched copy of the photo or scanned PDF
- A photo with all four page corners visible
- Even light with as little glare or motion blur as possible
- The intended page order when you add more than one input
Supported input includes JPG, JPEG, PNG, BMP, TIF, TIFF, WebP, HEIC, HEIF, and scanned PDF.
Correct the photographed page
Add the original
Open Fix Documents (BETA), add the photographed pages and arrange them in the intended PDF order.
You should see: each source once, in the order the finished document should use.
Check the page boundary
Open Adjust this page. Choose Keep as one page or Split into two pages. Use Find edges again, or move the four crop corners and adjust rotation and appearance yourself.
You should see: the complete page inside the boundary with no neighboring page fragment.
Compare both previews
Review Original and Cleaned. Drag to pan and zoom around the cursor to inspect corners, text, curves, shadows and lighting.
You should see: straighter edges and more even lighting without missing text or margins.
Create a clean copy
Save the default PDF as a standard A4 portrait or landscape page, then open the new file.
You should see: a new clean copy while every original photo or PDF remains unchanged.
Add searchable text only after the page looks right
OCR can create a searchable PDF after correction. The All languages option becomes the default once all 16 verified language packs are ready, and it detects the language of each text region locally. You can still choose one language for the entire document. If one page fails, successful pages remain searchable and the failed page stays as an image without searchable text. Inspect the result and compare important names and numbers with the page image.
If the cleaned page still looks wrong
Return to the original and check the page boundary first. If a book page is incomplete, photograph one full page with all four corners visible. If glare, blur or a hard shadow hides text, take the photo again with the camera parallel and the light spread more evenly. Correction can improve the appearance of captured detail, but it cannot recreate information that the camera missed.
Detection favors one complete page over a partly visible neighboring page. Two complete facing pages can be split into separate pages. If the boundary is unclear, the app keeps the full frame and asks you to review it.
PDF is the default output. Under Advanced options, Lossless PNG pages saves each corrected page as a separate image instead.
Document photo mistakes
- Photographing the page with a corner outside the frame
- Accepting a crop that includes part of the neighboring page
- Expecting correction or OCR to recover detail hidden by glare or blur
Cleaned PDF checkpoint
- The complete page appears once and no corner, text, signature, stamp or margin is cut off.
- The page is upright, proportions look natural and the image is not stretched to fill the intended A4 orientation.
- Faint writing and shaded areas remain readable when you zoom in.
- Search and copied text match the page image if OCR was added.
- If important detail disappeared, keep the untouched source and try a clearer photo.
Osenpa PDF Tools
Correct the photographed page, compare both previews and export a separate A4 PDF before adding searchable text or sharing the result.