Scanned, Screenshot or Encrypted PDFs: What Breaks

#Three file types, three different failures

A scan fails because it has no text, a screenshot fails because of resolution and aspect ratio, and an encrypted PDF fails because its content streams cannot be opened. The three are unrelated, and one set of steps will not fix them all.

Two questions identify which one you are holding: can you select the text inside, and does it ask for a password when you open it?

#Images to PDF: page size maths and text loss

When you use Images to PDF, every picture is converted to JPEG at quality 0.92 before being embedded. Two things follow: a transparent PNG gets a white background (so it will not turn black), and text screenshots lose a little sharpness.

PDF itself can embed images losslessly, but embedding JPEG keeps the file size of several dozen large photos manageable. That is a deliberate trade-off - pick original size and accept a bigger file if sharpness matters more.

The two page-size modes work differently:

• Original size: page = image pixels × 0.75, converting 96dpi pixels into PDF points. Long edges above 14400pt are scaled down to that limit, which is a constraint of the PDF format itself
• A4: 595.28 × 841.89 pt (210 × 297 mm) with 24pt margins (about 8.5mm), images scaled proportionally and centred - never cropped, never stretched

Processing is sequential, and the status bar reports "image 3 of 12". A picture that fails to decode is skipped and counted at the end, so one bad file does not abort the batch.

#A scan is an image, not text

Inside a PDF a scan is one full-page image with no text layer, so search, copy and edit are simply not possible. That is a property of the file, not a limitation of any tool.

The most reliable test needs nothing but your mouse:

• Open the PDF and try to select a sentence
• Nothing selects, it is a scan; text selects, it is a native PDF

Turning a scan into searchable text has exactly one route: OCR. That is recognition rather than conversion, and accuracy depends on camera angle, lighting and resolution - proofread anything important.

Merging a scan with a native PDF using PDF merge works fine, because merging concatenates pages at the document level. Afterwards, though, only the original native PDF is searchable.

#Encrypted PDFs must be skipped, not forced

A PDF with an open password keeps its content streams encrypted, so copying pages yields a document that will not open - which is why the correct behaviour is to skip it and say so. This is the situation most often mistaken for a broken tool.

The two tools respond slightly differently:

• PDF merge skips the encrypted file and reports how many were skipped
• PDF split refuses and explains why

There is one correct approach: remove the protection with the original password, then process. Without the password, give up - defeating the encryption is the whole point of the encryption.

#What merging and splitting do to file size

Neither merge nor split recompresses anything: page objects are copied as-is, so images, fonts and vectors are untouched. Quality is identical, and an extracted page is not a proportional slice of the size.

Three results that feel wrong until you know why:

• A single extracted page can be larger than expected, because the shared resources it references (fonts, images) live in the original file
• Merging two nearly identical PDFs gives roughly the sum, with no deduplication
• A merged file smaller than the sum usually means one input already carried redundancy or uncollected objects

Two behaviours in PDF split worth remembering: output follows the order you type, so 3,1,2 puts page 3 first, and repeated page numbers are kept once. A backwards range like 5-3 is read as 3-5.

It outputs one PDF per run. For one file per page, extract in batches or use qpdf --split-pages on the command line.

#Check a photo's metadata before you send it

A phone photo carries GPS accurate to a few metres, plus the device model and capture time, and the EXIF viewer shows all of it. Worth a look before sending a contract photo, a second-hand listing or a rental picture.

It reads three containers: the APP1 segment of JPEG, the eXIf chunk of PNG, and the EXIF chunk of WebP. HEIC and AVIF cannot be read; screenshots and images forwarded through chat apps have usually been stripped already, so an empty result is normal.

Two readings that look like errors but are not: an exposure time of 1/159 s comes from the camera storing an approximation of 1/160, and "35mm equivalent" is not the same as focal length - judge whether a lens is telephoto by the former.

#Which tool to use

Order matters: identify the file type before choosing a tool. Remove protection from encrypted files first, decide whether a scan needs OCR, and check that a screenshot has enough resolution to print.

#Frequently asked questions

#Text looks blurry after converting images to PDF. Can I avoid it?

There is some loss, because images are converted to JPEG at quality 0.92 first. To keep as much detail as possible choose original size rather than A4 - A4 rescales, which means small images grow and large ones shrink, and interpolation itself costs sharpness. The source resolution sets the ceiling: a low-resolution screenshot only gets softer.

#Why was one of my PDFs skipped during merge?

It has an open password. Encrypted content streams stay encrypted, so copying pages produces a file that will not open, and the tool skips it and reports the count. Remove the protection with the original password and merge again.

#Can I search text in a scanned PDF?

No, not without OCR. A scan is one full-page image with no character data, so search, copy and edit have nothing to work with. OCR is text recognition, not a format change, and its output needs proofreading.

#Why is the split file not the size I expected?

Because splitting does not recompress: the extracted pages are copied as-is and still reference shared resources such as fonts and images. Size depends on what those pages actually reference, not on how many pages you took out.

#Is a phone photo the same as a scan?

For processing purposes yes - both are an image placed in a PDF with no text layer. The difference is quality: phone photos suffer perspective distortion, shadows and uneven lighting, so OCR accuracy is noticeably lower than from a flatbed scanner.