Convert Picture PDF to Text When the Pages Are Phone Photos
By Schema LawPublished Updated
To convert a picture PDF to text, run the file through OCR, because a PDF built from phone photos holds only images of pages and no characters. Upload the PDF, let every page be read, compare the parsed pages with the pictures, then export plain text, Word or a searchable PDF.
A photo PDF is harder work than a file from an office scanner, and the reason is simple. A scanner treats every sheet the same way. A person holding a phone over a kitchen table does not. The sections here show how to recognize the file, how to convert it, and how to decide whether to ask for better pictures.
What is a picture PDF and why has it no text?
A picture PDF is a PDF in which each page is a photograph of a sheet of paper. Somebody photographed the pages with a phone, and an app or a print-to-PDF command wrapped those photos in one file. The file stores pixels only, so there is nothing to select, search or copy.
The PDF wrapper misleads people. A file ending in .pdf looks like a finished document, and the person who sent it may believe they have "scanned" the paper. What they did is put photographs in an envelope. The envelope keeps the pages in order and lets one attachment travel in place of twelve, and that is all it adds.
The post on why you can't copy text from a PDF explains why an image-only file has no selectable text and gives a quick test for it. What matters for a photo PDF is what sits inside the pictures, because OCR can only read what the camera recorded.
A photographed page differs from a scanned one in ways that affect recognition:
- The page is not flat to the lens. A phone held at an angle records a page that is wider at one end than the other, so lines of type slope and letters at the far edge are smaller.
- Light falls unevenly. A lamp to one side, a window behind the phone or the shadow of a hand leaves part of the page darker than the rest.
- The picture holds more than the page. A desk, a bedspread, a thumb or the next sheet in the stack can appear around the edges.
- Sharpness varies. A camera focuses once for the whole frame, and a moving hand softens every letter.
"Text" can mean three different results, so settle which one you want before you start. Plain text is the words alone. A Word file is words you can edit and format. A searchable PDF keeps the photographs as they are and adds a hidden layer of text behind them. One conversion can give you all three.
Choose the searchable PDF when each photograph shows a page of a client's lease you will keep as documentary evidence. The text layer lets you find a clause, and the pictures stay on the page for checking. Leave the original photo PDF unchanged so authentication and the chain of custody refer to the file the client actually sent.
How do you tell a PDF was made from phone photos?
Open the PDF and look at the page edges, the page sizes and the file size. A photo PDF shows surroundings beyond the paper, pages of differing shapes, and several megabytes for a handful of pages. A scanner file has uniform pages, square corners and nothing visible outside the sheet.
Run through these signs with the file open.
| Sign | Where to look | What a phone photo shows |
|---|---|---|
| Surroundings | The margin of each page | A table top, fabric, fingers or another sheet beside the paper |
| Shape of the paper | The four corners | A page that narrows toward the top or leans to one side |
| Page dimensions | Document properties, or the size shown when you hover near a page corner in some readers | Sizes that are not a standard paper size, or that change from page to page |
| Orientation | Scroll through quickly | Upright pages mixed with sideways ones |
| Lighting | Compare the top and bottom of a page | A gradient from light to dark, or a shadow with a soft edge |
| File size | The file's details in your file manager | Megabytes per page where a typed PDF would be kilobytes |
| Producer | The document properties panel | A phone app or a print-to-PDF component named as the creator, if the field is filled in |
No single sign settles it, and a scanning app that crops and flattens the page hides several of them. Three or four together are enough to treat the file as photographs.
Why bother identifying it?
Knowing the file is a photo PDF changes what you do next. You check every page, where with a feeder scan you might sample. You expect some pages to convert well and others badly. And you know that the person who sent it still has the paper, or at least the phone, so a second attempt is possible.
A worked example
A fictional client, Dana Whitcombe, emails a file named lease photos.pdf. It has nine pages and is 31 MB. Page 1 shows a lease on a wooden table with a coffee mug at the corner. Pages 4 and 5 are sideways. Page 7 is taller and narrower than the rest. The document properties give no standard paper size. That is four signs, and you treat the file as nine separate photographs that happen to share a wrapper.
Converting a picture PDF to text from upload to export
Converting a picture PDF to text takes six steps: keep the file as received, check its size and page count, fix page-level faults, upload a copy, read the parsed pages against the photographs, and export the format you need. Do them in that order so you never work on the only copy.
- Save the PDF as it arrived. Put the attachment in the matter folder under its original name and do not open it for editing. Work on a duplicate.
- Check size and page count. Photo PDFs are heavy. Schema Law takes PDF, PNG or JPEG files of up to 20 MB, and the plans differ in pages per document: 3 on Free, 250 on Solo and 1,000 on Team. A nine-page file at 31 MB is over the size limit even though it is short. The post on converting a large scanned PDF to Word online covers shrinking or splitting an oversized file.
- Turn and tidy the pages. Rotate sideways pages upright in your PDF reader and save the duplicate. Note any page that is cropped or holds two sheets.
- Upload and wait for the pages to be read. Sign in, add the file from your browser, and let the document finish. The three steps on the homepage describe the same flow.
- Read every parsed page beside its photograph. With a photo PDF, look at each page. Names, dates, amounts and section numbers come first. Schema Law's homepage says recognition is strongest on clean print and that nothing marks a passage as uncertain, so the comparison is your only check.
- Export. Choose plain text for the words alone, Word for editing, or searchable PDF to keep the photographs with text behind them. Every plan, Free included, exports all three.
Which export for which job?
| You need to | Export | What to know |
|---|---|---|
| Quote a paragraph in an email or a memo | Plain text | Layout is dropped; see what plain text keeps and what it drops |
| Revise or reuse the wording | Word | Check headings, numbering and tables before you edit |
| File the document so search finds it | Searchable PDF | The photographs stay as they are, shadows and table top included |
A searchable PDF made from photographs still looks like photographs. If the file is going to someone who expects a tidy document, the Word export or a fresh capture serves better than a text layer behind a picture of a kitchen table.
For the comparison in step 5, the post on why OCR output is full of errors explains how to judge a page of recognized text and where mistakes hide.
Why do pages in one photo PDF convert so unevenly?
Pages in one photo PDF convert unevenly because each page is a separate photograph taken under its own conditions. The distance, angle, light and focus changed every time the sender lifted the phone. OCR reads page 2 without knowing that page 1 was sharp, so quality rises and falls page by page.
A feeder scanner applies one resolution and one light source to a whole stack. A hand-held phone applies neither. Five things shift between shots:
- Distance. A photo taken from farther away spends fewer pixels on each letter. Small print and footnotes suffer first.
- Angle. A tilted phone makes the far edge of the page smaller and less sharp than the near edge.
- Light. The sender moves, a cloud passes, the overhead light reflects off a glossy page. One page is bright and the next is gray.
- Focus and movement. A page photographed in a hurry is soft all over. A page photographed while the camera was still hunting for focus is sharp in one area only.
- The paper. Page 1 lies flat. Page 6, still stapled, curls up at the corner and bends the lines of type.
Keep a page log
A page log turns an uneven file into something you can describe and rely on in parts. Make one row per page as you compare the parsed text with the photograph.
| Page | Photograph | Parsed text | Action |
|---|---|---|---|
| 1 | Sharp, flat, even light | Reads correctly | Use |
| 2 | Shadow across lower third | Two rent figures wrong | Correct by hand from the image |
| 3 | Taken from far away | Footnote unreadable | Ask for a retake |
| 4 | Sideways | Read after rotating | Use |
| 5 | Curled corner | Last line of each paragraph garbled | Ask for a retake |
The log does three jobs. It tells you which pages need a retake, so your request to the sender is specific. It tells a colleague which parts of the text have been checked. And it stops a clean-looking export from being trusted more than the worst page deserves.
Does a bad page spoil the good ones?
No. Recognition runs page by page, so a blurred page 3 does nothing to the text of page 1. The risk is a human one: someone searches the export, finds nothing, and concludes a term is absent when it sits on the page that failed. Record the weak pages where the next reader will see them, such as the file name or the first line of a cover note.
Sideways, cropped and double-page pictures inside a PDF
Fix sideways pages by rotating them before upload, treat cropped pages as incomplete, and decide how to handle a picture that holds two sheets. These are faults of single pages, and each one is cheaper to deal with before conversion than in the exported text.
Sideways and upside-down pages
Rotate the page in your PDF reader until the text reads normally, then save the duplicate. Tools differ in whether they turn pages on their own, and Schema Law's site makes no statement about it, so do the turning yourself or test one sideways page first. The post on preparing a PDF for OCR covers orientation along with blank and duplicate pages.
Check for a second problem while you rotate. A sender who turned the phone for one page may have photographed that page from a different distance, so look at its sharpness too.
Cropped pages
A page cut off by the edge of the photo is missing text, and no software can restore words the camera never recorded. Look at all four edges of every page. Signs of a crop include a sentence that stops mid-word at the right margin, a page number missing where the other pages have one, and a signature block with no signature line. The post on scanning legal size documents has a section on spotting cut-off text in a scan you were sent.
Mark a cropped page in your log as incomplete, and say so wherever the converted text is used. A lease clause quoted from a page whose last line is missing should carry that warning.
Two sheets in one picture
A picture of an open booklet or of two sheets side by side puts two pages of text into one PDF page. Two things can go wrong. The type is smaller, because the camera stood back to fit both sheets. And the reading order may run straight across both sheets, weaving the left page into the right. The post on whether OCR can handle two-column documents explains reading order and how to check it.
If your PDF editor can crop, duplicate the page and crop one copy to each sheet before upload. If it cannot, convert the page as it is, then read the result closely for order. Either way, note in your log that PDF page 6 holds paper pages 6 and 7, so later page references match the document and not the file.
Pictures in the wrong order
Photographs taken out of sequence stay out of sequence in the PDF. Read the printed page numbers or the paragraph numbering from page to page, and reorder the duplicate before conversion. A converted text that jumps from section 4 to section 9 is easy to miss once the page images are out of sight.
Asking the sender for better pictures or working with the PDF you have
Ask for better pictures when a page that matters is cropped, blurred or unreadable to you, and the sender still has the paper. Work with the PDF you have when every page is legible to your own eye, when the paper is gone, or when the deadline will not wait.
A simple rule helps: if you cannot read a word in the photograph at full zoom, OCR has nothing to recognize either, and only a new picture will fix it. If you can read the page comfortably and the software stumbled, careful proofreading may be quicker than a second round with the sender.
| Situation | Better course |
|---|---|
| Text is cut off at an edge | Ask for a retake of that page |
| A page is too blurred or dark for you to read | Ask for a retake of that page |
| Pages are legible but shadowed or skewed | Convert, then correct against the image |
| The sender no longer has the original | Work with the file and record its limits |
| The file is needed today and the pages are readable | Convert now, request retakes for the weak pages in parallel |
| The document's appearance may matter later | Ask for the paper itself, or a scan of it |
Making the request
A vague request gets the same pictures back. Name the pages, say what was wrong, and say how to send the new ones. For example, to the fictional Ms. Whitcombe: "Pages 3 and 5 of the lease came through too blurred to read. Could you photograph those two pages again, one page per picture, with the whole sheet in the frame, and email them at full size?" The guide on how to photograph documents for your lawyer is written for the person holding the phone, and you can send its link with the request.
When retakes arrive
Do not overwrite anything. Keep the first PDF as received, save the new pictures under their own names, and build a working copy that swaps in the better pages. Record which pages came from which message. If the text will be quoted or relied on, the file should show where each page image came from and when.
Then run the conversion again on the working copy and update the page log. The second pass is quicker, because you already know which pages to read first. The pricing page sets out the page limits for each plan if the repeat run takes you past the free allowance.