Turn photos, scans and screenshots into one PDF: choose the page size, orientation and margins, see the pages, and keep every JPEG at its original quality.
A phone photo is 4:3; A4 paper is roughly 1:1.41. Something has to give, here are the three options.
Almost every country uses A4. The US, Canada and parts of Latin America use Letter, which is a little wider and shorter. Pick the one your reader will print on, the wrong one gets scaled or cropped at the printer.
Not printing at all? Choose Fit to image - each page matches its photo exactly, with no white edges.
A PDF can carry a JPEG untouched: its bytes go straight into the file under the DCTDecode filter, so no quality is lost. Everything else is decoded to pixels and squeezed with FlateDecode, the same compression ZIP uses. Below is the PDF this page just wrote, checked against your originals and taken apart object by object.
1 0 obj<< /Type /Catalog /Pages 2 0 R>>endobj…Before compressing, FlateDecode can store each byte as the difference from a guess, the pixel to the left, above, or a blend (Paeth). Smooth areas turn into runs of zeros that compress well. PDF borrowed these five filters from PNG. Here they run on your first image, every row of it, compressed by your browser.
/Filter /FlateDecode/DecodeParms << /Predictor 15 % PNG filters, chosen per row /Colors 3 % R, G, B /BitsPerComponent 8 /Columns width>>Adaptive picks, for every row, the filter whose output has the smallest sum of absolute values, the heuristic the PNG specification suggests. This page uses it for every non-JPEG image it writes.Eight filters can sit on an image stream. Version numbers are the PDF version that introduced each one; check ISO 32000 for the exact wording before relying on them.
DCTDecodeused by this page1.0lossyJPEG (baseline and progressive)A JPEG file can be copied in unchanged; EXIF orientation is ignored, so the page has to rotate it.FlateDecodeused by this page1.2losslesszlib-compressed pixels or textWith /Predictor 10-15 it uses PNG’s row filters; 2 is TIFF’s horizontal predictor.LZWDecodeused by this page1.0losslessolder lossless imagesSuperseded by Flate; not allowed in PDF/A.RunLengthDecodeused by this page1.0losslessruns of identical bytesSimple; good only for flat graphics.CCITTFaxDecodeused by this page1.0lossless1-bit black-and-white scansGroup 3 and 4 fax coding; still common in scanned archives.JBIG2Decodeused by this page1.4either1-bit scans, shared symbolsLossy “pattern matching” mode can swap similar glyphs, the 2013 Xerox digit-substitution problem.JPXDecodeused by this page1.5eitherJPEG 2000Carries its own colour space; supports 16-bit and lossless.(none) + /SMaskused by this page1.4-alpha channel as a separate greyscale imageTransparency; forbidden in PDF/A-1, allowed from PDF/A-2.Before your images go in, or after the PDF comes out.