Phone photos to a clean searchable PDF
Goal
Turn a dozen phone photos of paper pages into one straight, upright, searchable and reasonably small PDF.
You photographed a lease, page by page, on the kitchen table. Some shots are a few degrees crooked, two are sideways, and each photo is 4 MB. You want one file you can search with Ctrl+F and email.
Starting state
- Photos as JPG/PNG (or TIFF, BMP, WEBP), named in page order, shot mostly straight on. Deskew only rotates; it cannot fix perspective (a photo taken at an angle) or curved book pages.
- Up to 10 files and about 30 MB per Merger run; more photos means two rounds.
- You know the document's language.
Chain
Merger → Rotation → Deskew → Searchable → Compressor
photos sideways small add text shrink last
to PDF pages up tilt out layer
Step by step
1. Merger (images to one PDF)
Drop the photos, drag them into page order, keep Mode: Sequential, choose a Page size if you want one (A4 by default), press MERGE. Why not Converter: Converter handles each file separately; you want one PDF. See Merger.
2. Rotation
Choose Selected pages, list the sideways pages (for example 3,7), pick 90° or 270°: look at where the top of the text sits. Why before Deskew: Deskew handles only small tilts (about 15 degrees) and OCR reads sideways pages badly. See Rotation.
3. Deskew
Upload the PDF and press the button; there are no options. It measures every page, corrects those tilted more than about 0.3° and leaves straight pages alone. Look at the previews. Be aware: the output is page images, so any text layer is gone. See Deskew.
4. Searchable
Choose the Document language. Keep Deskew image on (it is harmless after step 3 and helps stubborn pages), tick Remove background/watermark for shaded or stained paper, leave Source photo is sideways off (you fixed that), keep Output format on Searchable PDF. Press Transform. Why after Deskew: level text recognises better. See Searchable.
5. Compressor
Choose Medium, press COMPRESS. Why last: Searchable adds a text layer and enlarges the file, and compression is lossy. If pictures look too soft, redo from the Searchable result with Low. See Compressor.
Why this order
Pixels first, then text: you must get the pages upright (Rotation) and level (Deskew) before recognition reads them, and you must read them before you shrink the pictures it reads from. Rotation before Deskew because the second only handles small angles.
Expected result
One PDF that looks like your photos but whose pages are level, whose text can be searched and copied, and whose size is a fraction of the sum of the photos.
Check it worked
- Page count equals the number of photos.
- Pages are upright and level when you scroll.
- Ctrl+F finds a word you can see; selecting text works.
- Spot-check a name or number: recognition is an estimate, and wrong words look right.
- The size panel from Compressor shows a clear saving.
Variations
- Only the words: skip the PDF and send JPG/PNG straight to OCR Image.
- Very long (hundreds of pages): Searchable may refuse; split with Splitter, process, merge back.
- Tilted photos with perspective: retake them flat from above; no app here fixes perspective.
- Wrong words: try Searchable's Correct OCR errors with AI and check the result; it can change words.
- Add chapters: Add TOC before Compressor.
What can go wrong
- Deskew left a page crooked: tilt under 0.3°, too little text, or more than about 15°.
- Text not selectable after Deskew: expected; step 4 fixes it.
- File got bigger after Deskew: images were re-encoded; Compressor reduces it.
- Garbage words: wrong language or low resolution.
- Skipping step 2 and ticking the sideways option: it applies to all pages, not only some.