There is a small decision that hides inside almost every scan, and most people never make it on purpose. The app picks for them, or they pick once and never think again, and then months later they open a file and something is wrong: a signature that should have been blue is a flat gray smudge, or a simple one-page letter has somehow become a four-megabyte monster. The color mode you scan in is not a cosmetic choice. It changes what the file knows, how much room it takes up, and whether the page can still prove what it was meant to prove.
Here is the idea worth understanding deeply: color, grayscale, and black-and-white are not three flavors of the same picture. They are three different amounts of information per pixel, and the right one depends entirely on what the page is carrying.
What "black and white" actually means
When a scanner says black and white, it almost never means a dark gray photo. It means bitonal, sometimes called 1-bit. Every single dot on the page is forced to be one of two values: pure black or pure white. Nothing in between. To get there, the software runs a process called thresholding — it looks at each pixel, asks "is this darker or lighter than my cutoff?" and snaps it to one side.
For crisp printed text on clean white paper, this is wonderful. Letterforms have hard edges anyway, and removing the in-between grays makes them sharper and the file tiny. This is why fax machines and decades of document archives used bitonal images: a page of text becomes a small, clean, high-contrast pattern that compresses beautifully.
The danger is everything that lives in the in-between. A pencil note, a faint thermal receipt, a photograph, a watercolor — thresholding doesn't soften these, it executes them. A pixel that was a delicate gray becomes either black or gone. If the page has any tonal subtlety, bitonal scanning will flatten it into something blotchy and broken.
What grayscale keeps that black and white throws away
Grayscale, usually 8-bit, gives every pixel one of 256 shades from black to white. That sounds modest, but 256 steps is enough for the human eye to read a smooth gradient as continuous. Suddenly a pencil sketch holds its softness, a faded letter keeps its ghostly middle tones, and a photocopy with uneven shadows stays readable instead of crumbling.
Grayscale is the quiet workhorse for anything that isn't pristine print but doesn't depend on color: handwritten notes, old typewritten pages, receipts that have started to fade, book pages with a yellowed cast. You lose hue, but you keep depth. And depth is what makes a difficult page legible.
It's also worth knowing that most OCR engines convert your image to grayscale or bitonal internally before they read it, because they care about the shape of the ink, not its color. So if your only goal is turning a clean printed page into searchable text, color buys you almost nothing for the recognition step itself.
What color is really for
Full color is typically 24-bit — roughly sixteen million possible values per pixel, split across red, green, and blue. That is an enormous amount of information, and you should spend it deliberately, only when color is carrying meaning.
The clearest case is ink that proves something. There is a long-standing convention, especially in legal and banking contexts, of signing in blue ink precisely so that an original can be told apart from a black-and-white photocopy. Scan that signature in bitonal and you have erased the one feature that distinguished it. A blue official stamp, a red "PAID" mark, a notary seal, a highlighter stripe a colleague added to flag a clause — all of these are information, and all of them vanish or turn into meaningless gray the moment you drop color.
Color also matters when the page is partly or wholly an image: a brochure, a chart whose categories are color-coded, a child's drawing, a printed photograph, a passport or ID page where the security design and portrait are the point. Here the color is the content, and anything less throws away the document's reason for existing.
The file-size tradeoff nobody warns you about
This is where the choice becomes practical rather than philosophical. The three modes differ enormously in how much storage they eat.
A bitonal page is the lightest by a wide margin, because one bit per pixel compresses extremely well — long runs of identical white space collapse to almost nothing. Grayscale is heavier. Color is heaviest of all, often many times the size of the same page in grayscale, especially at high resolution.
Multiply that across a filing cabinet's worth of paper and it stops being trivial. Scanning a hundred pages of plain printed text in full color doesn't make them more truthful; it just makes them slow to sync, slow to send, and bulky to back up. The instinct "color is the safe maximum" is the same instinct that makes people film everything in 4K and never watch it. More data is not more memory. It is just more data.
A simple way to decide
You don't need a rulebook. You need one question, asked at the moment you point the camera: does color or fine shading carry meaning on this page?
If the page is clean printed text and you mainly want it searchable and small — bitonal black and white is honest and efficient. If the page has handwriting, fading, soft pencil, or uneven tone but no meaningful color — grayscale preserves the legibility without the bloat. If the page has a signature in colored ink, a stamp, a highlight, a chart, a photo, or any visual that color is — color, without hesitation.
There's also a gentle default if you genuinely can't tell and the document matters: grayscale for ordinary records, color for anything you might one day need to prove is an original. You can always make a color scan smaller later. You can never put back a hue you threw away at capture.
Why getting this right at capture matters
The reason this decision deserves a moment of thought is that it is mostly irreversible in the direction that hurts. Converting a color scan to grayscale or black-and-white afterward is trivial. Going the other way is impossible — there is no button that invents the blue back into a signature once it's been flattened to black. The information either entered the file at capture or it didn't. Every other edit you'll ever make to that document assumes the color decision was already made correctly.
That is also why it pays to use a scanner that shows you what it's doing rather than guessing silently. LumenScan lets you choose color, grayscale, or black-and-white per scan and see the result before you commit, and because its OCR and processing run entirely on your device, that blue-ink signature or sealed certificate never has to leave your phone to be read or stored. The color you choose, and the page itself, stay yours.
If you'd like a scanner that treats each page as a decision rather than a default, you can find LumenScan at https://lumenscan.lumenlabs.works — and the next time you scan, you'll know exactly which of the three to reach for, and why.