Ink or paper, nothing between
A photo of a typed letter is mostly paper. Written as a colour JPEG it carries three channels of that paper, and even in grey it carries eight bits of it for every dot. In black and white each dot is one bit — ink or paper — and the page is coded the way a fax machine codes it: CCITT Group 4 describes a whole run of white as a few bits however long the run is, and describes each line by how it differs from the line above. For a page of text that is the smallest honest file there is, and the letters come out with hard edges instead of the faint ringing a JPEG leaves around them.
A worked example from our own test set: a 1654 × 2339 photo of a typed declaration, lit unevenly from one corner, put on A4 with 10 mm margins. Copied in colour the PDF weighed 865,692 bytes; redrawn in grey, 714,247 bytes; in black and white, 59,930 bytes — about a twelfth of the grey file — written as Group 4 at 221 DPI across the page, with the shadowed corner coming out as clean white paper. With a second page carrying a red rubber-stamp seal the two-page file was 120,150 bytes; the builder flagged the seal page before saving, and the seal printed solid black.
How the page decides what is ink
A single cut-off for the whole page fails on a phone photo: the corner nearest the window is brighter than the corner under your hand, so one threshold either loses the text in the shadow or turns the shadow black. Each dot here is compared with its own neighbourhood instead — a window about an eighth of an inch across at the page’s resolution. A dot clearly darker than the paper around it is ink; flat paper, even grey paper, stays paper. That is Sauvola’s method, and it is why the background of a black-and-white page comes out white rather than blotchy.
- Normal suits printed and typed pages.
- Darker keeps faint marks — a light ballpoint signature, a pale photocopy — at the cost of some speckle.
- Lighter drops grain and show-through from the back of thin paper.
Each page is written both ways, as Group 4 and as packed bits through Flate, and the smaller one is kept. On text Group 4 almost always wins; the receipt after saving says which was used for every page.
When to use grey instead
Black and white is for text. A photograph on the page becomes a pattern of dots, pencil shading becomes speckle, and a coloured seal or stamp keeps only its darkness. If a portal wants a certificate with its seal legible, or the document carries a photo of the applicant, choose Grey on the Quality step — the size routes, JPG to PDF under 100 KB and under 200 KB, start in grey for that reason. For a crooked or shadowed photo of a page, square it up first with scan picture to PDF’s document mode, then come back to black and white; photographing a document properly helps more than any setting.
Questions about JPG to PDF in black and white
What is the difference between grey and black and white?
Grey keeps every shade between paper and ink, eight bits a pixel, written as JPEG. Black and white decides, pixel by pixel, whether it is ink or paper — one bit a pixel — and writes the page with the coding fax machines use. Grey is right for photos and pencil; black and white is right for typed and printed text, where it gives the smallest file with the hardest edges.
Will a stamp or seal survive?
Not always, and the builder says so before you save. A red or blue seal becomes whatever its darkness decides: a dark stamp turns solid, a pale one can disappear. When a photo has clearly coloured marks, the Quality step names the page and suggests Grey for that document instead.
Can I make an existing PDF black and white here?
No. This page makes a new PDF from photos and images of documents. It does not open an existing PDF and recolour it.
Is the text in the PDF searchable?
No. Each page is a picture of the page in black and white; there is no text layer, so the words cannot be selected or searched. The file is smaller and sharper, not machine-readable.
Are my photos uploaded?
No. The thresholding and the Group 4 coding run in this tab on your device. Nothing is sent anywhere, and the page keeps working with no connection once it has loaded.