The number in that box is a label, not a measurement
The instruction is the same everywhere. Open the image size dialog, make sure the resample box is unchecked, type 300 into the resolution field, save. Guides built around that one step rank at the top of this search, and the online converters offer 72, 300 and 600 as buttons so you do not even have to find the dialog.
It works, in the narrow sense that the file afterwards says 300. What almost nothing explains is that the pixels are untouched. Not improved slightly, not improved invisibly: identical, every one of them, byte for byte. You changed a note attached to the file about how large it would like to be printed, and a note is not detail.
So the useful version of this question is not how to raise the number. It is how to work out whether your file has enough pixels for the print you want, which is one multiplication and takes less time than finding the dialog did.
What the resolution field is for
An image file is a grid of pixels and a set of notes about that grid. The pixel grid is the picture. One of the notes records an intended physical density, in dots to the inch, and its only job is to tell a printer or a page layout how large to place the picture when nobody has said otherwise.
That note is a default, and like any default it is overridden the moment something downstream specifies a size. Place the same file in a layout and drag it to the width you want and the note has no further effect. Send it to a print shop with a size on the order and the same is true.
This is why re-tagging a file feels like it should help and never does. The grid is the evidence, the tag is a preference, and the tag is the only thing the standard instruction changes.
The multiplication that actually decides it
A print is a physical object with a width in inches. A density is how many dots the printer will lay down across each of those inches. Multiply the two and you have the number of pixels you need across. That is the whole calculation and it runs in either direction.
Going forward: a print eight inches across, at a density of 300, needs a file about 2400 pixels across. Going backward, which is the more useful direction: open the file, read its pixel width, divide by the density you have been asked for, and the answer is the largest size it prints at that density. A file that comes out wider than the print you wanted is already sufficient and nothing needs doing to it.
Notice what is not in that sum. The resolution tag does not appear anywhere. It cannot, because the printer is told the physical size by the job and the pixel count by the file, and those two facts are enough to determine the density. The tag is a suggestion that has already been superseded.
The checkbox is the entire decision
Every tool that offers to change DPI is really offering two completely different operations, and which one you get is decided by a single checkbox labelled resample, or resize, or scale depending on the software. The guides tell you to leave it off, which is correct advice for the wrong reason: it is correct because you probably do not want what the other setting does, not because the result is harmless.
- Resample off. The pixel grid is untouched and only the note changes. The file size on disk stays the same, because nothing was added. This is the operation every tutorial on this search describes.
- Resample on. The software invents new pixels between the ones you have, by averaging the neighbours, so the grid genuinely grows and the file on disk gets larger. Nothing about the subject was recovered, because nothing new was ever recorded.
- The giveaway is the file size. If the saved file weighs the same as the original, you re-tagged it. If it weighs much more, pixels were added, and added pixels from averaging are smooth rather than sharp.
- A model-based upscaler is a third thing again, and the only one that adds something that was not in the original. It proposes what the detail probably was and draws it in, which is useful and is not the same as recovery.
Run the test yourself in a minute
This is worth doing once, because reading it is less convincing than seeing it. Take any photo and export it twice from the same original, once tagged 72 and once tagged 300, with resample off both times.
Then compare the two files. The pixel dimensions are the same in both. The file sizes are the same in both. Open them side by side at the same zoom and there is nothing to see, because there is nothing different to see. Inspect the metadata and exactly one field differs between them, and it is the field you set.
Print both at the same physical width and they come out indistinguishable, which is the part that settles the argument: the printer was told the width by the print job, worked out the density it needed from the pixels available, and never consulted the tag at all.
What that test rules out is the hope that the number was doing something subtle. It was not. If the print was soft before, it is soft for a reason the tag cannot reach.
Most of the time your file is already fine
The ordinary case for this search is somebody who has been told by a printer, a form, or a submission portal that an image must be 300 dpi, and who has a file from a recent phone or camera. Those files are large. Do the division and they usually print comfortably at a normal photographic size without anything being done to them.
In that situation the honest answer is that there is nothing to fix, and re-tagging the file to say 300 is a reasonable thing to do anyway: it costs nothing, it satisfies a checker that reads the field, and it does not change the picture. Just be clear with yourself about which of those two things you did.
The case that is genuinely a problem looks different. A small image saved off a website years ago, a screenshot, a thumbnail somebody emailed, a crop from a group photo. Divide the pixel width by the density and the answer comes back smaller than a postcard. No tag fixes that, because the pixels are the shortfall.
When you really are short of pixels
If the sum says your file is too small for the print, the only thing that helps is more pixels, and there are exactly two places to get them.
The first is upstream, and it is almost always the better answer. If a print, a negative, a slide or the original camera file still exists, that object holds detail no small copy of it carries, and recapturing it gives you resolution that is genuinely there rather than proposed. People routinely upscale a low quality phone snap of a print that is sitting in a drawer in the next room.
The second is an upscaler, which estimates the missing detail from everything it has seen before. Texture, fabric, foliage and brickwork come back convincingly because they behave statistically. Faces and text do not: a face at low resolution leaves the answer far less constrained than it looks, and the result is a clear confident face that belongs to nobody. Crisp and wrong is worse than blurred and honest, and faces are the part of the picture people are looking at.
Which of those two you need is the same question the resolution guide works through at length, and it is worth reading before you commit to an enlargement you intend to hang on a wall.
What this means for Unblur Photo
Unblur Photo adds pixels and sharpens what is there, and it does not set a DPI tag, because a tag is a field in a file rather than a thing an enhancer can improve. If a portal is rejecting your file for its resolution field, that is a one-click fix in any image editor and this app is the wrong tool for it.
What it is for is the other half of the problem: a file that is genuinely short of detail, where the print will be soft no matter what the field says. The output is larger than the input and carries more information than the input, and that information is inferred rather than recovered, which matters for a face and does not matter for a wall.
The guides on improving photo quality and on unpixelating an image cover where that inference is trustworthy. Read them before you print something that has to establish a fact rather than fill a frame.
More guides
Questions
Does increasing DPI improve image quality?
No. With resampling off, changing the resolution field edits one note in the file and leaves every pixel identical, which you can confirm by checking that the saved file weighs the same as the original. Quality is set by how many pixels you have and how much real detail they carry, neither of which the field touches.
How do I convert an image to 300 dpi?
In any image editor, open the image size dialog, switch resampling off, and set the resolution field to 300. Online converters do the same thing in a browser. Be clear that this satisfies a checker reading the field and changes nothing about the picture, so if the print was going to be soft it still will be.
How many pixels do I need for a 300 dpi print?
Multiply the width of the print in inches by 300. Eight inches across needs about 2400 pixels across. Run it the other way to answer the question you usually have: divide your file pixel width by 300 and that is the widest print it supports at that density.
Why does my printer say the resolution is too low when the file says 300 dpi?
Because the checker is doing the division rather than reading the tag. It takes the physical size on the job, compares it with the pixel dimensions of the file, and reports the density that combination actually produces. A tag claiming 300 on a file with too few pixels does not survive that calculation.
Is DPI the same as resolution?
They get used interchangeably and they are not the same. Pixel dimensions are how much picture exists. Dots per inch is how densely those pixels are laid down on paper, which only has a value once a physical size is chosen. One is a property of the file and the other is a property of a print.
Should I upscale the image or just change the DPI?
Do the sum first. If the pixel count already covers the print size, change the field if a form demands it and leave the picture alone. If the pixel count falls short, the field cannot help and you need more pixels, ideally by recapturing the original and otherwise by upscaling, which estimates detail rather than recovering it.