This came about on a group I am part of. The group consists of amateur, novice, hobby, and professional artists who all utilize a print-on-demand service. While often times there are discussions about copyright, fanart, and sales, this person was saying that she had a shop open and someone said her images looked “blurry”.

Upon some discussion, it was realized that the shop owner was making images, saving them as a JPG file format image, and then scaling up as necessary. Most well practiced artists already know this, but here’s so advice for “new” artists, or ones who may want to use a print-on-demand service:

NO NO NO NO NO

Whenever you’re working with a source file, you want to make sure that source file is as crisp, clean, and as large as possible. So let’s say you sketch something and you want to digitize it in some way. Maybe you want to color it, or you want to print it on a t-shirt, or something like that. You want to make sure you get the best photo possible. That means if you’re scanning it, you want to scan at around 300 to 600 DPI to get a really nice and clean version of your image. When you’ve done that, you want to save this as either a high-quality lossless JPG, or a different lossless file format like PNG or TIFF.

This is your source file. Keep it safe. Keep it separate.

Then move on to coloring or editing as you need to with your file. If you’re using Photoshop, you’re going to want to probably use a PSD file (at 300 DPI MINIMUM) and work on it until you’re done. If you’re anything like me, you’ll have 6 or 7 versions and various PSD files with your edited work before the final thing is done. This is your edited file, but your final source file. Keep it safe. Keep it separate.

Finally, create your flattened JPG, PNG, or otherwise for web or other usage. A third and final file, or more if you’re like me and create full resolution flattened files, low resolution (web) flattened files, and files for your website/portfolio.

That’s file structure right there. But let’s talk about the root of the issue here.

Let’s say you don’t have a good source file. Maybe you’ve only got an itty-bitty photo someone took once on a flip phone that you just love. You want to blow it up. The answer is to size it up, right? No big deal.

Not really.

If it’s your only option, sure, go ahead and do that. Blowing up a photo from a negative is different than blowing up a digital photo, though, so you might wonder how one can have different outcomes. The thing is that digitally you can only produce something as good as your source file. Here’s why:

If the pixels don’t exist, then they have to come from somewhere

Let’s talk about math for a second. Just like a 10’x10′ area has a total square footage of 100 square feet, a 10px by 10px image has 100 total pixels. So that means that a 1000×1000 px file has 1,000,000 pixels total. A 2000×2000 pixel image has 4,000,000 total pixels. Wow, that’s a lot of pixels!

If you’re taking a file that is 1000×1000, and you’re scaling it to 2000×2000, that’s an extra 3,000,000 pixels inside of the file that need to be accounted for. A computer is not going to drop transparent pixels and call it a day — that wouldn’t look very nice. Pretty much what a computer has to do when you ask it to make a picture bigger is that you’re asking it to fill in all the missing pixels for you. In programs like Photoshop, there are certain algorithms built in to fill in those pixels a certain way. Some are optimized for scaling up or down images with sharp edges. Others are meant for smooth gradient-like effects. The skinny of it is that the computer will guess based on surrounding pixels on what to put in between the pixels of the original image. The more pixels you’re asking the computer to guess, the less true to the original image it will become, and the blurrier the end result will get. Of course this only counts with raster images and not vector images, which can be scaled infinitely without loss of quality, but that’s a whole different discussion.

Here’s a visual I made to display it:

Each image (except the first one) was a relatively small image (800×1000 pixels) scaled up to the size of the original source image using the various scaling algorithms. They’re not terrible, but the bigger I scale the image, the worse they’ll get.

The tl;dr version of this is: start big and scale down after! If you plan on making prints or anything else, you’ll have a much easier time making huge images and scaling them down after the fact. That way you won’t get blurry, pixelated products — or any product complaints about it.