How to increase resolution of a image for a badge printer

A badge printer is a small piece of paper with a strict count. Upscale 2× and measure the result before you jump to 4×.

My sister is printing reunion badges on a borrowed dye-sub printer, and her note started with the phrase she had typed into a search box: how to increase resolution of a image. The attachment was a passport-style crop, head and shoulders, plain wall, sent from her phone's mail app. The file is 180 by 220 pixels and 22 KB. The printer software threw a low-resolution warning and offered to "enhance," which is how I got the call.

The badge stock is a standard card, 3.37 by 2.13 inches. The photo window in her template is 1.1 inches wide and 1.4 inches tall. At 300 pixels per inch, that window wants about 330 by 420 pixels. Her attachment is short of that on both sides. It is not short by a factor of ten. It is short by a bit less than half on the long side. That distinction matters. A file that is slightly small can be printed small and look acceptable, or it can be replaced by the original crop that was sitting on the phone the whole time.

I asked her to look at the phone, not at the email she had already sent herself. The mail app had a size setting, and it was on medium. Medium, for this app and this photo, meant the 180 by 220 file. The crop she had made the night before was still in the camera roll at 1320 by 1650. Same head, same gray wall, same slight tilt of the chin. Face height in the thumbnail is about 150 pixels. Face height in the camera-roll crop is about 1100 pixels. The badge only needs a few hundred.

TL;DR. The mail app sent a 180 by 220 thumbnail of a passport-style crop, and the badge window needs about 330 by 420 pixels at 300 per inch. That thumbnail can be enlarged a little and will look soft on the card. The phone still had a 1320 by 1650 crop, which is more than the printer needs, so the real fix was a second email with the large file attached as actual size. I did not upscale the thumbnail once the large file appeared. A driver setting was also shrinking files tagged 72 ppi, and I set the inches myself.

The thumbnail, measured before anyone enhances it

I saved the attachment without opening it in a viewer that "helps." The pixel size is 180 by 220. The JPEG is aggressive: the gray wall breaks into blocks, and the hairline has a dark fringe about 1 pixel wide that is compression, not hair. Eyes are roughly 9 pixels across. A mole on her left cheek, which I know is there because I have seen it for years, is a single darker pixel. That mole is my test target. Any pass that turns it into a neat circle with a soft edge has drawn a mole, not kept one.

The crop is passport-style in the loose sense. Head centered, a little shoulder, no smile, wall behind. It is not a government portrait and I am not making it one. The reunion badge needs a recognizable head at a doorway, viewed from about 2 feet. At 2 feet, a 1.4 inch photo is a small object. Softness that would ruin an 8 by 10 on the mantel can pass on a lanyard. I still want the large file, because the large file is free and the thumbnail is a choice the mail app made.

The warning in the printer software quoted 72 dpi. It was reading a metadata tag, not measuring the card. A 180 by 220 file tagged 72 ppi would claim to be about 2.5 by 3 inches, which is larger than the photo window, so the driver wanted to scale it down and then complained that the result would be soft. Both statements can be true at once if you let the tag pick the inches. I do not let the tag pick the inches. I pick the inches from the template: 1.1 by 1.4.

The second email, which ended the job for most people

She re-sent the camera-roll crop as actual size. It arrived at 1320 by 1650, 1.4 MB, same mole as a small cluster of pixels rather than one dot. I placed it in the template at 1.1 by 1.4 inches and let the layout scale it down. Scaling down is an ordinary resample. The printer receives more pixels than the window needs, and the driver throws some away. Thrown-away pixels from a real photo are a luxury. Invented pixels from a thumbnail are a patch.

I did not run an upscaler on the 1320 file. There was nothing to gain. The cheek already had texture. The hairline fringe from the thumbnail was gone, because that fringe lived in the small JPEG, not on her head. The large JPEG has its own compression, mild, and I left it. A denoise would have sanded the mole's cluster into a dot again.

Three other volunteers had done the same medium-email trick. I wrote a one-paragraph instruction: open the photo, share, mail, actual size, and do not screenshot the badge preview. Two of them followed it. Their files were 1000 to 1600 pixels on the long side, all fine. The third had deleted the camera roll copy. For her, the 180 by 220 attachment was the only file. That one card is the upscale story. The others are a file-handling story. How to increase resolution of a image, in this office, mostly meant how to stop the mail app from being the photographer.

The one card that had to be enlarged

For the deleted original I used a 2 times pass on the AI image upscaler, after a light cleanup of the wall blocks. Output: 360 by 440. That just covers the 330 by 420 window. I stopped there. A 4 times file would have been 720 by 880, and the extra size would have been guesswork past what the card can show. I compared the mole. In the thumbnail it is one dark pixel. In the 2 times file it is a soft four-pixel smudge, slightly rounder than I wanted, still in the right place. I accepted it because the alternative was the raw thumbnail printed soft, and the smudge was not a new feature in a new place.

Eyes got rounder. I will say that plainly. The 9-pixel irises became smoother circles. At badge size I cannot see iris pattern anyway, because there was no iris pattern in 9 pixels. The risk was a change of expression. I looked at the mouth. It stayed closed and level. If the pass had added a smile, I would have printed the thumbnail untouched. A smile she did not wear is a worse error than a soft edge.

I also ran the thumbnail through the plain resize in the layout, bicubic, to 330 by 420, with no model. Side by side on the card proof, the bicubic was softer and more honest, and the 2 times model was a bit crisper and a bit more "portrait studio." My sister preferred the model for the one card, knowing the original was gone. I preferred the bicubic for the wall and the model for the hair. We shipped the model file, and I noted the preference in the folder so I do not pretend I was neutral.

The enhance photo quality route is what I would use if the large file had been soft from focus, rather than small from an email setting. This thumbnail was small. The large files were sharp. Mixing a general enhance on top of a 2 times upscale, on a face this tiny, stacked two guesses. I tried it on a copy, saw the mole become a beauty mark with a highlight, and deleted the copy.

The mole rule

If I know a mark on a relative's face, I use it as the stop. If a pass invents a highlight the mark never had, the pass is done. Badge size will hide a lot of softness. It will not hide a beauty mark that makes a sibling say the photo looks like someone else.

What the driver was doing with the ppi tag

The borrowed printer's software has a checkbox, on by default, that says it will fit images to the frame using the file's resolution tag. For the 1320 by 1650 crop, the phone had tagged 72 ppi. A naive reading of that tag says the image is about 18 by 23 inches. Fit-to-frame then shrinks an 18 inch picture into 1.4 inches, which is a heavy downsample, and the preview looked muddy even though the pixels were fine. The mud was the preview resampling twice, once in the fit and once in the on-screen proof.

I turned the checkbox off and set the photo frame to 1.1 by 1.4 inches in the template. The same 1320 file snapped sharp in the preview. Nothing about the JPEG changed. The interpretation of it changed. I have watched people upscale a file to "satisfy" this kind of warning when the warning was a tag. The upscale makes a bigger JPEG for the driver to shrink again. It is a circle.

I also set the output of the driver to the printer's native panel. This model prints the card at 300 by 300 of its own pixels per inch. I do not feed it 600 and hope. Extra resolution above the panel gets filtered by the driver anyway. The useful question is whether the face, inside 1.4 inches, has enough real samples to survive that filter. The 1100-pixel face does. The 150-pixel face barely does, and only after we accept softness.

A table of the reunion files

I kept one row per source so the next reunion does not start from memory.

Source	Pixels	Face height	What I did
Mail medium thumbnail	180 by 220	about 150 px	Upscaled only when this was the only file
Camera-roll crop	1320 by 1650	about 1100 px	Placed on the card, no upscale
Volunteer range	1000 to 1600 long side	enough for 1.4 in	Actual-size email, then the template
2 times from thumbnail	360 by 440	about 300 px	One card, mole stayed a smudge
Bicubic to the window	330 by 420	soft but honest	Proof only, not the shipped card
Driver tag mistake	72 ppi tag on a large file	looked muddy	Set inches in the template, ignored the tag

The last row is the one I would put in a manual. Pixel count and the ppi tag answer different questions. The tag is a suggestion about inches. The count is the information. How to increase resolution of a image is often a request to edit the suggestion, or to go get the count that already exists.

Checks at the laptop next to the printer

We had one afternoon and a stack of blank cards. The list lived on a sticky note on the laptop lid.

• Read the attachment's pixel size before reading the driver's warning.
• Ask if the phone still has the crop, and ask for actual size, not medium.
• Measure the photo window in inches off the template, not off the preview.
• Multiply those inches by 300 for this printer, and treat that as the budget.
• Use a known mark on the face as a stop if you must upscale.
• Turn off fit-to-frame when it is obeying a 72 ppi tag.
• Set the name as type under the photo. Do not pull letters out of the picture.
• Pro one proof card and look at it in the hallway, not only on the screen.

The hallway matters. The office monitor is bright and large. The reunion doorway is neither. A proof that looks soft on the monitor can look fine on a lanyard, and a proof that looks sharp on the monitor can still have a color cast under the hallway bulb. We caught a green cast that way, from a profile the driver applied only at print time. I turned the profile off, reprinted one card, and the wall went back to gray. That had nothing to do with resolution and it would have ruined the stack.

A drawing I can show without a volunteer's face

Badge photos are faces, and I am not putting a reunion portrait on a public page. The stand-in I use is a generated drawing: round head, teal rectangle, a panel with a yellow circle, a dark rectangle, three tan bars, a small green shape on a brown floor. The edges are clean because the drawing was made of flat colors, not because a model recovered them.

I show it when someone asks what "enough pixels" looks like on a simple edge. The yellow circle stays a circle. The tan bars stay separate. Nothing has to be guessed. A 180 pixel headshot is not that drawing. The eye is 9 pixels and the mole is one. If I blur the drawing on purpose and shrink it, I can demonstrate a soft edge. I still should not pretend the demonstration is her face. The drawing keeps the lesson, and the proof card keeps the verdict.

The clean drawing also keeps me from practicing on the volunteers. I will test a wild setting on the drawing first. If the yellow circle grows a ring or a second outline, I do not need a person's cheek to tell me the setting is wrong.

Color, crop, and the name line

The large crop had a slightly warm wall, from indoor bulbs. The thumbnail looked cooler, because the small JPEG had crushed the color. When both were printed with the same driver settings, the large file matched the hallway better. I did not "correct" the large file toward the thumbnail. The thumbnail was the worse record.

Chin tilt was about 3 degrees. The passport-style instinct is to rotate the head until the eyes are level. I rotated the large file by that 3 degrees, cropped a hair tighter, and left a margin of wall so the template mask, not my crop, would be the final edge. Rotating a 180 pixel thumbnail is harsher. The resample from a 3 degree turn on so few pixels softens the eyes again. Another reason to start from 1320.

The name under each photo is set in a plain sans at 14 points, black, from a spreadsheet of names she typed. One volunteer had a name tag in the original snapshot, and the thumbnail had included a blurry scrap of it. I cropped that scrap out. I will not sharpen junk letters into a name that might be wrong. The spreadsheet is the name. The photo is the face. They meet in the layout.

Background removal came up. The template already masks to a rectangle with a thin white rule. Cutting the wall away and floating the head made the badges look like stickers, and the hair fringe from the thumbnail became a halo against the card's white. I refused the cutout for the upscaled card and I refused it for the sharp cards too, so the stack would match. Consistency across 40 badges beat a clever mask on one of them.

What "increase the resolution" meant in the search

The search phrase she used has a small grammatical wobble, and I am leaving it intact because it is the phrase that arrived with the file. People who type how to increase resolution of a image are often holding a file that a program has already resized, and they want the program to undo its own resize. Sometimes the undo is possible because the earlier file still exists. Sometimes the undo is a modest enlarge and a smaller print. Rarely is it a return to detail the sensor never kept.

Her camera-roll crop is the undo. The medium email is the damage. I want that sentence to be boring. Boring file handling saved 39 cards from an upscaler. The one upscaled card is labeled in the spreadsheet with a note, so if she reprints next year she knows which face was guessed.

If the phone had also been gone, and the 180 by 220 file were the archive, I would print the bicubic version at the true window size and I would not promise eyes. I would tell the volunteer. We were lucky that only one person was in that spot, and she laughed about it, which is not a technical requirement but it is how the stack got finished without a fight.

Proof cards and the stack

We printed four proofs: thumbnail bicubic, thumbnail 2 times, large file with the bad checkbox, large file with inches set by hand. I laid them on a table under the hallway bulb and I also stepped back 6 feet, which is the distance from the registration table to the door. At 6 feet all four were recognizable. At arm's length the checkbox disaster was muddy, the thumbnail bicubic was soft, the 2 times card was a little plastic in the cheeks, and the large file was obviously the photograph.

The stack used the large-file setup. The one 2 times card went into the stack at the right alphabet position. I kept the proofs in an envelope so nobody "cleans up" the folder later and leaves only the plastic card as the example of a good job.

Ink or ribbon, in this case a dye-sub ribbon, has its own sharpness limit. Fine hair against a gray wall prints as a soft boundary even from a sharp file. I do not sharpen the large file to fight the ribbon. Sharpening put a pale edge on the hair in a test strip, about a pixel in the print file, and it looked like lint on the card. The ribbon will be slightly soft. That is the printer. Leave it.

Folder names, because reprints happen

reunion-thumb-180.jpg, reunion-roll-1320.jpg, reunion-2x-360.jpg, reunion-proof-checkbox.jpg, reunion-proof-fixed.jpg. The template file sits beside them. The spreadsheet of names is there too. I wrote the window size, 1.1 by 1.4, and the printer's 300, on the first line of a notes file. Next year the borrowed printer may be a different model. The inches of the card will probably be the same. The pixels we need will follow from the inches and whatever panel the new printer actually has.

I also saved the mail app's medium file untouched, because it is the evidence. If someone later says the upscaler "created" resolution, I can open 180 by 220 next to 1320 by 1650 and show that the resolution was in the camera roll. The model was a fallback for a deleted roll, and it is named like a fallback.

Neighboring notes, different problems

How to enhance resolution of image is the note I send when the file is soft as well as small. How to fix resolution of a photo covers the tag and the inches, the mistake our driver made. Improve photo resolution is the short version for someone who has not found the original yet. This reunion is specifically a thumbnail that should not have been the file, plus one person who had to live with the thumbnail.

If you are holding a passport-style crop and a badge template, measure the window in inches, multiply by the printer's real pixels per inch, and compare that budget to the attachment. Then look at the phone. How to increase resolution of a image is a strange little sentence, and the straightest reading is the one that helped us: increase the number of real pixels by attaching the picture that already has them.

The card she wears in the sample pack on my desk is the large-file proof. The mole is a cluster. The wall is gray. The name is type. The thumbnail lives in the folder and not on a lanyard, except for the one person who already knows why.