Enlarge a Photo Without Losing Quality: What Is Possible

You have a small photo and you need a bigger one, for a print, a screen or a form. The honest answer to "can I enlarge it without losing quality?" is: you can make it bigger without making it worse, but you cannot add detail the camera never recorded. This guide explains what enlarging does, when Fast mode or AI mode is the better choice, and how to work out the size you need.

What a photo is made of

A photo is a grid of tiny coloured squares called pixels. A photo that is 800 pixels wide and 600 pixels high has 480,000 of them. Everything you can see in the photo comes from those squares.

When you enlarge a photo, the computer has to fill in new pixels. Making it 2× doubles the width and the height, so the picture has 4 times as many pixels. Making it 4× gives 16 times as many. The new pixels are worked out from the ones you already have.

For example, an 800 × 600 photo made 4× bigger becomes 3200 × 2400 pixels. That is 7,680,000 pixels (7.7 megapixels) made from 480,000.

What enlarging can and cannot do

It can:

• give a small picture enough pixels for a bigger screen or a bigger print
• make edges look clean instead of blocky
• with AI mode, clean up the blocky marks that heavy JPEG compression leaves

It cannot:

• bring back a face, a number plate or a line of small text that was too small to see in the original
• fix a photo that is out of focus or blurred by movement
• make a photo as sharp as one taken at the bigger size

Enlarging only spreads the detail you have over more pixels. A method can make that look cleaner, and AI can add believable texture, but it is a good guess, not a recording of what was really there.

Fast or AI?

SafeResize offers two modes.

Fast is a careful, sharp enlargement. It works on modern phones and computers. It does not add new detail, so nothing in the picture changes except its size. In our test it took about 2.5 seconds for a 768 × 512 photo on a laptop, and it is slower on a phone.

AI uses a neural network, Real-ESRGAN, an open-source model made for enlarging pictures. It runs on your device's graphics chip. It takes longer than Fast, and it is only offered on devices whose graphics chip can run it. The first time you use it, the page downloads a 2.3 MB file from saferesize.com, once, and shows the size before it starts. At 2× and 3× the AI result is blended with a plain enlargement (60% AI at 2×, and 80% AI at 3×). At 4× it is all AI.

In our tests we shrank 8 sharp, well-lit photos, enlarged them back to the original size with each mode, and compared them with the originals. AI came out as close to the original as Fast, or closer, in every test at 2×, 3× and 4×, with the biggest gain on photos that had been compressed as JPEGs. The difference is small, not dramatic.

It does not always win. On a dark, noisy, low-light photo, AI can score lower than Fast, because it removes the noise and the original still has it. The AI result looks smoother, not necessarily worse, but it is not closer to the original.

Choose Fast when the photo is already decent and you want a bigger copy that is still true to the original, for a photo with faces or with text you need to read, or on a phone with a low battery.

Choose AI for a small, blocky or heavily compressed picture, such as an old download, a thumbnail or a low-quality scan, and look at the faces and the text afterwards. AI can make skin and surfaces look smoother than they were, and it can change small letters.

You do not have to decide first. Make it in one mode, then press the button for the other mode and look at both.

How big do you need it?

Work from the pixels you need, then see how far the photo has to be enlarged.

For a print, about 300 pixels for every inch gives a sharp result. For a big print that people see from farther away, about 150 pixels for every inch is often enough. To find the pixels, multiply the size in inches by 300 (or 150).

For a screen, count the pixels of the screen. A Full HD screen is 1920 × 1080 pixels. A 4K screen is 3840 × 2160 pixels, which is 8,294,400 pixels.

Now compare. A 1000 × 750 photo made 4× is 4000 × 3000 pixels. That is enough for an 8 × 10 inch print at 300 per inch (2400 × 3000), with pixels to spare. The same photo made 2× is 2000 × 1500 pixels, which gives about 190 pixels for every inch on an 8 × 10 print.

The size limit

SafeResize makes pictures of up to 16,777,216 pixels, which is 4096 × 4096 for a square picture. No side is longer than 8,192 pixels. The limit is the same on every device and was chosen to suit phones. At 300 pixels per inch, 4096 pixels is about 13.7 inches, and at 150 it is about 27 inches.

What this means in practice:

• A photo of up to about 1 megapixel (for example 1024 × 1024 or 1280 × 800) can go the full 4×.
• A bigger photo gets a smaller step. A 12-megapixel phone photo (4000 × 3000) can only be made about 1.18× bigger, to 4720 × 3540 pixels.
• A photo that already has 16.8 megapixels or more cannot be enlarged, and the tool says so.

If the choice would be too big, the tool lowers the enlargement by itself and tells you. Also, 4× is the most in one go. For an 8K picture (7680 × 4320) you would need about 33 million pixels, which is over the limit.

Steps in SafeResize

1. Open Upscale an image and choose the photo. The photo stays on your device.
2. Pick how much bigger to make it. Photos up to 1024 pixels on the long side start at 4×, and bigger ones start at 2×. If your device can run AI mode, pick Fast or AI.
3. Compare before and after with the slider, or press "Full size" to see real pixels.
4. Save it as a JPG or a PNG. For a JPG you can set the quality, or keep the file under a size in KB or MB. Files are named like photo-4x.jpg.

There are pages with the size already chosen: upscale an image 4× and upscale an image to 4K. On enlarge a photo the tool picks the size for you.

Tips for the best result

• Start from the original file. A photo that was sent through a chat app, or saved from a web page, was already compressed. Every round of JPEG saving loses a little detail, and enlarging makes those marks bigger.
• Enlarge once. Making a photo 2× and then 2× again adds the marks of each round. Choose the size you need in one go.
• Save a PNG if you will edit it more. JPG makes a smaller file, and PNG keeps every pixel as it is.
• Check the faces and the text. Zoom in with the "Full size" view before you save, especially after AI mode.
• Crop first. If the picture has to be cropped, do that before you enlarge, so no time is spent on parts you will throw away.

When not to enlarge

Do not enlarge or "improve" a passport or visa photo. Official photos must be unedited, and a photo that is too small should be taken again at a proper size. The guide on taking a passport photo at home explains how. Also be careful with any picture that has to show what is really there, such as evidence or a document: an enlarged or AI-changed copy may not match the original.

The file gets bigger too

A picture with 16 times the pixels makes a much bigger file. If a website has a limit, tick "Keep the file under a size" when you save a JPG, and the picture is made a little smaller if it has to be. To go the other way, see KB, pixels and cm explained.

Privacy

Both modes run on your own device, and the photo is never uploaded. In AI mode the page fetches the model file, a file that is the same for everyone. The photo is not part of that request.