OCR finishes on iPhones that cannot run the AI engine
With the engine set to Auto, a page that is a design rather than a scan (a poster, a slide, an infographic) is read by PP-OCR. On some iPhones that engine cannot start, and the whole run failed with an error about memory, once after fifty minutes of work on a 408-page file. Auto now reads those pages with Tesseract and finishes.
Very large results need less memory
Most tools made a second copy of the finished PDF before handing it to you. With a few gigabytes of photos in JPG to PDF, that copy was the one the browser could not fit, and a result over 2 GB was refused outright. The copy is gone, and a result over 2 GB is accepted.
A blank page no longer stops the run
Fix Page Size, N-up, Booklet and the other tools that place your pages on new sheets failed with "Invalid header in flate stream" when a file held a blank page saved in a way some programs use. That page now comes through blank, like the rest of the file.
All changes in version 2026.11.19-01
- FixedID Card Composer, JPG to PDF, Image to PDF, CBZ to PDF: a photo that failed with "Invalid JPEG" converts
- FixedWhen the PDF engine fails to download, the page reloads once to fetch it, instead of failing every file until you reload
- FixedProtected PDFs: when the component that unlocks them is blocked or fails to download, a second copy is tried, and if that fails too the error says what to check
- FixedOCR PDF: an engine you picked that cannot start tells you which ones will