Scan Handwriting Into Text: Capture Settings That Help
By Schema LawPublished Updated
To scan handwriting into text, capture the original page in grayscale or color at 300 dpi or more, flat and uncropped, so that thin and faint pen strokes survive. Recognition software can only work with the strokes the image holds. A harsh black and white scan or a shadowed photo removes them before reading starts.
Why does the scan matter more for handwriting than for print?
The scan matters more for handwriting because pen strokes are thinner, fainter and less regular than printed type, and each letter appears in only one version. A printed word with a damaged letter can still be recognized from the rest of the font. A handwritten word that loses a stroke becomes a different word.
Printed type is built to survive copying. Its strokes are evenly black and the same width from top to bottom, and every "e" on the page is identical. Handwriting has none of that margin. Pressure changes within one letter. The upstroke that joins two letters may be a third as dark as the downstroke beside it. A pen skips for a millimeter and leaves a gap.
The parts of a handwritten letter that tell it apart from its neighbors are usually the smallest marks on the page.
| Small feature | What it distinguishes | What happens when the scan drops it |
|---|---|---|
| The dot over an i | i from a short l or the first stroke of u | "minute" loses its shape; "in" may become "m" |
| The crossbar of a t | t from l | "tie" reads as "lie"; "mitt" reads as "mill" |
| The closing stroke at the top of an a | a from u, and o from an open c | "cat" and "cut" become a guess |
| The small loop of an e | e from c and from an undotted i | "these" can come back as "those" or "thcse" |
| The bar of a 7 or the flag of a 1 | 7 from 1 | A date or an amount changes |
| A faint joining stroke | Where one letter ends | Two letters merge into one |
A scanner or camera can lose these marks in three ways. It can decide they are too light to count as ink. It can blur them into the paper through heavy compression. Or it can cover them with something darker, such as a shadow or a ruled line. Each of the capture choices in this post is about one of those three.
There is a second reason the scan carries more weight. With print, a careful reader can often spot a recognition error because the result is not a word. With handwriting, a wrong reading tends to be a real word, so errors hide. A better image is the cheapest way to have fewer of them. What software can make of the strokes once it has them is a separate matter, covered in the post on OCR for handwritten notes.
Color, grayscale or black and white for handwritten pages?
Scan handwritten pages in grayscale as the default, in color when the ink or the paper is colored, and never in pure black and white. Black and white mode makes the scanner decide which marks are ink. With handwriting it often decides wrongly, and the lost strokes cannot be recovered from the file.
| What is on the page | Mode | Reason |
|---|---|---|
| Black or dark blue pen on white paper | Grayscale | Light strokes are kept as light gray |
| Pencil | Grayscale, or color on tinted paper | Graphite is gray, not black, and sits near the cut-off |
| Blue or red pen on a form printed in black | Color | The entries can be told apart from the printing |
| Any ink on yellow, pink or green paper | Color | The background stays a color and does not turn to gray grain |
| Two inks on one page, such as later corrections in red | Color | Who wrote what, and when, stays visible |
| Notes with highlighter over them | Color | Highlighter scans as a dark band in black and white |
Three scanner behaviors undo a sensible choice of mode, so check for them.
Automatic color detection. Many office devices pick the mode for each page. A sheet of pale pencil on white paper can be judged "black and white" and scanned that way. If you scan handwritten notes to text in batches, set the mode by hand.
Text enhancement presets. Settings with names like "text", "sharpen text" or "remove background" raise contrast to make print crisp. On handwriting they push faint strokes to white. Use the "photo" or "mixed" preset if the device offers one, or switch enhancement off.
Heavy compression. A "small file" or "compact PDF" preset smears thin lines into the paper. Choose the lighter compression setting and accept a larger file.
On resolution, 300 dpi is the working floor for handwriting of ordinary size. Go to 400 dpi for small or cramped writing, such as notes squeezed into a calendar square. The general guide to scanner settings that make OCR work covers resolution, duplex and file format for a whole file room, and the post on grayscale versus black and white scanning explains the modes themselves.
Watch the file size when you move up to color. A long color scan grows quickly, and upload limits apply. Schema Law accepts a PDF, PNG or JPEG of up to 20 MB, so a thick notebook may need to go up in parts.
Pencil, blue ink and felt tip under the scanner
Each writing instrument fails in its own way under a scanner. Pencil is too light, ballpoint is uneven, felt tip is too heavy, and colored inks change brightness when the color is thrown away. Knowing which one is on the page tells you which setting to change before you scan.
Pencil
Pencil is the hardest common instrument. Graphite is gray and slightly shiny, the line is thin, and writers press lightly. In grayscale a pencil note is often only a little darker than the paper. Scan it in grayscale or color, and if the scanner has a brightness or exposure control, darken the scan by one step and compare. Do not rub or "clean up" a smudged pencil page before scanning. Smudges are gray haze that grayscale can separate from the strokes, and rubbing spreads them.
Ballpoint
Ballpoint is dark where the writer pressed and pale where the pen glided, and it skips on glossy or greasy paper. Expect broken letters at the start of a line, where the ball had not yet picked up ink. Grayscale holds the pale parts. Ballpoint also dents the paper, and on a thin sheet the dents from the other side can show as faint ridges. Those are harmless unless a black and white scan turns them into marks.
Blue, red and green ink
Colored ink is the case where color mode earns its file size. A light blue ballpoint looks clear to your eye because it is a different color from the paper. Convert it to gray and it may be only a shade darker than the page. Red ink behaves unpredictably: some scanners render it nearly black in grayscale and others render it pale. If corrections, initials or dates are in a second color, scan in color so they stay distinct from the first writer's ink.
Gel pen, felt tip and marker
Dark gel ink scans well and is the easiest handwriting to capture. Felt tip and marker have the opposite problem from pencil: the strokes are so wide that the loops of "e", "a" and "o" fill in, and small writing turns into a row of blots. More resolution helps here, because it puts more pixels inside each closed loop. Marker also soaks through. Check the back of the sheet, and if writing from the reverse shows on the front, rescan on the flatbed with a sheet of black paper behind the page.
Carbonless form copies
The pink and yellow copies from a multi-part form carry a faint blue or purple trace on colored paper. They combine the two worst conditions. Scan them in color at 400 dpi, and if the white top copy exists anywhere, ask for it.
| Instrument | Main risk | First thing to try |
|---|---|---|
| Pencil | Strokes too light | Grayscale, one step darker |
| Ballpoint | Skips and pale strokes | Grayscale, no text enhancement |
| Light blue or red ink | Fades when color is removed | Color |
| Felt tip or marker | Loops fill in; bleed-through | Higher resolution; check the reverse |
| Carbonless copy | Faint trace on tinted paper | Color at 400 dpi; look for the top copy |
Yellow legal pads, ruled lines and spiral notebooks
Paper causes as much trouble as ink. A yellow legal pad lowers contrast, ruled lines cut through the letters, and spiral or perforated edges jam feeders and cast shadows. Scan legal-pad pages in color, keep the ruling in the image, and put torn-out or bound pages on the flatbed.
Yellow paper
In grayscale, yellow becomes a medium-light gray, which narrows the gap between paper and ink. In black and white, parts of the yellow can fall on the dark side of the cut-off and print as speckle across the page. Color mode keeps the yellow as yellow, and software can then separate it from blue or black ink by hue as well as by brightness.
A yellow pad often holds a durable power of attorney rather than a memo. You check the principal's name, the person named as fiduciary or guardian, and any line about a conservatorship. Light blue ink on that yellow sheet is exactly where a black and white scan drops a word you still need.
Ruled lines and the margin rule
The blue rules and the red margin line on a pad run straight through the writing. A rule under a word can look like an underline. A rule through the tails of "g", "y" and "p" can join one line of writing to the next. The red margin line can be read as a tall letter "l" or the digit 1 at the start of every line.
Some scanners offer a "color dropout" setting that removes one color, usually to erase form lines. Be careful with it on handwriting. Dropping blue removes the rules and also any blue ink. Dropping red removes the margin line and also any red corrections. If you try it, keep a full-color scan as the record and treat the dropout version as a working copy.
Graph paper is a harder version of the same problem, since the grid crosses every letter in both directions. Scan it in color and expect more correction.
Spiral notebooks, steno pads and perforations
A page torn from a spiral notebook has a ragged strip of holes down one side. In a sheet feeder that edge catches, skews the page or tears further. Use the flatbed. If you may not remove pages, because the notebook is someone's original record, open it flat on the glass, scan one page at a time and press gently near the binding so the writing by the spiral stays in focus. The holes scan as a column of dark ovals, and software may read them as zeros or the letter O. That is one more reason not to crop tightly: you want to see where the margin junk came from.
Writing on both sides
Thin pad paper lets the reverse side show through. Place a sheet of black paper behind the page on the flatbed. A dark backing makes the show-through blend into an evenly darker page, where a white backing reflects light back through the sheet and outlines every stroke on the other side. Scan both sides, in order, even when the back holds only a phone number.
Sticky notes, envelopes and scraps
Small pieces travel badly through a feeder. Lay them on the flatbed, several to a scan if you like, with space between them. Scan a sticky note once where it was found and once on its own, so the record shows both what it said and what it covered.
Can you copy handwritten text from a phone photo or a whiteboard picture?
Yes. A sharp, evenly lit photo of handwriting can be sent for recognition like a scan, and the text can then be copied out. The photo needs enough pixels across each letter, no shadow over the writing and no glare. Whiteboard pictures add reflections and faded marker to that list.
People asking how to copy handwritten text from an image usually have one of three files.
A photo you can retake. Retake it properly. The steps for taking a photo that OCR can read apply in full. For handwriting, add two habits. Get closer than you would for print, photographing a dense page in two overlapping halves if you must, because pen strokes need more pixels than type of the same height. And avoid the scanning app's black and white filter, which treats pencil and light ink the way a black and white scanner does.
A photo someone sent you. Messaging apps shrink pictures. A client's texted photo of a handwritten note may be a fraction of the size the phone captured. Ask for the original file by email, or ask the client to share it at "actual size" or "original quality". If the page still exists, ask for a new photo with the whole sheet in the frame.
A picture of a whiteboard or flip chart. Dry-erase marker on a glossy board gives you glare from ceiling lights, colors of very different darkness, half-erased earlier writing and a strong keystone shape when the photo was taken from a seat. If the board is still there, stand square to it, turn off the flash, and photograph it in sections with overlap. If all you have is the meeting photo, crop to the board, expect yellow and light green marker to drop out, and plan to type the gaps from the image.
To copy handwriting text from an image once you have a usable file, upload the JPEG or PNG, let the pages be read, and export plain text or Word. Then compare what you copied with the photo before it goes anywhere. Other picture formats, such as HEIC from an iPhone or a TIFF from a copier, are covered in the post on scanned image to text.
Screenshots of handwriting, for example a tablet note shared in a chat, are a different animal. They are already digital and usually sharp, but small. Ask for an export from the note app at full size if you can.
Inspecting a handwriting scan before you send it for recognition
Inspect a handwriting scan for two minutes before recognition, with the paper still on the desk. Zoom in on the faintest writing, check the edges, and confirm every side was captured. A fault you find now costs one rescan. A fault you find after correcting ten pages of text costs the ten pages.
Work through the image in this order.
- Find the faintest line on the page and zoom to 200%. It may be a pencil note, a date squeezed in a corner or the last line before the pen ran dry. If you can read it on screen as easily as on paper, the mode and brightness are right.
- Look for the small marks. Check that i-dots, t-bars, commas and decimal points are present. If periods have vanished, the scan is too light or too compressed.
- Check the loops. In felt-tip writing, look at a few "e" and "a" shapes. If they are solid blobs on screen but open on paper, rescan at a higher resolution.
- Check all four edges. Writing that runs to the edge of a pad is common. No line should be cut at the right margin or the bottom.
- Count sides. The number of images should equal the number of written sides. Note blank backs so nobody later wonders whether they were missed.
- Check the order and the orientation. Notes written with the pad turned sideways should be rotated upright as a separate working copy, with the scan as made kept too.
- Look for shadows and show-through. A gray wedge near a notebook's binding or ghost writing from the reverse means a rescan on the flatbed with a dark backing sheet.
- Check the file. Confirm the format and size your converter takes before you upload.
| What you see on screen | Likely cause | Fix |
|---|---|---|
| Pencil lines broken into dashes | Black and white mode, or enhancement on | Rescan in grayscale, enhancement off |
| Yellow page covered in gray grain | Legal pad scanned without color | Rescan in color |
| Halos or blocky smears around strokes | Compression too high | Rescan with lighter compression |
| Blue entries missing entirely | Color dropout or a copier's blue filter | Turn dropout off; rescan in color |
| Letters stretched near one edge | Page lifted or curled | Flatten and rescan on the flatbed |
Once the image passes, convert it. The post on how to convert handwriting to text takes over from here, with the steps for judging the draft and correcting it against the page. If you want to try the capture on a real page first, the three steps on the Schema Law homepage show what an upload involves, and the free plan takes a document of up to three pages a day. The homepage says plainly that recognition is weaker on handwriting than on clean print, so judge by your own pages.
Keep the paper until someone with authority over the file says otherwise. A scan made with care is a good working record, but it is the handwriting on the sheet that a client, an opposing party or a court may one day want to look at.