How to Upscale a Small Photo Without It Looking Fake
Everyone has a photo that's too small: an old picture, a thumbnail saved from a website, an image that looks fine on a phone but turns blocky when enlarged. Upscaling can help, but it's easy to misuse and to expect too much from. This guide explains what upscaling actually does, how many pixels you really need, and how to get a result that looks natural rather than fake.
Why enlarged photos look bad
A digital photo is a grid of pixels. Enlarge it and the software has to invent the new pixels in between. An ordinary resize does that by blending neighbouring pixels, which makes the image bigger but also softer, because no new detail is created. The bigger you go, the blurrier it looks.
What AI upscaling does differently
An AI upscaler has been trained on many pairs of low- and high-resolution images, so it has learned what edges, skin, hair, foliage and fabric typically look like at higher resolution. When it enlarges your photo, it reconstructs plausible detail instead of just blending pixels. On small, blurry or heavily compressed images this usually looks noticeably sharper. The important word is "plausible": the detail is a convincing guess, not information recovered from the original scene.
How many pixels do you actually need?
For print, the usual target is 300 dots per inch, so the pixels you need equal the print size in inches times 300:
| Print size | Pixels needed at 300 dpi |
|---|---|
| 4 × 6 in | 1,200 × 1,800 |
| 5 × 7 in | 1,500 × 2,100 |
| 8 × 10 in | 2,400 × 3,000 |
| A4 (8.27 × 11.69 in) | 2,480 × 3,508 |
Larger prints that are viewed from a distance, such as posters, often look fine at 150–200 dpi. For screens, pixels matter more than print size. Our guide to pixels, KB and DPI explains how these numbers relate.
Know the limit of the tool
The Quikbench AI Image Upscaler works on images up to about 2 megapixels (roughly 1,600 × 1,250 pixels). A larger photo is first reduced to that size, so upscaling a big photo gives little or no benefit: a 12-megapixel phone photo shrunk and then enlarged 2× comes back smaller than the original. The tool warns you before you spend credits.
That makes it ideal for exactly the images that need help:
- An old scan or web image around 640 × 480, which at 4× becomes 2,560 × 1,920.
- A small crop. If you need to enlarge just a face or a detail from a big photo, crop it first, then upscale the crop.
Choosing 2× or 4×
Pick the smallest scale that gets you the pixels you need. 2× costs 1 credit and 4× costs 2, and a bigger enlargement means more reconstructed, less real, detail. If 2× is enough for your print size, stop there.
Keeping it natural
- Start from the best version you have. The more real detail goes in, the better the result.
- Don't upscale twice. Run the original once at the scale you need.
- Check faces and text. Enlarged faces can look slightly different from the real ones, and small text can be redrawn imperfectly. Compare with your original, and keep the original file.
- For a blurry face, the AI Face Restorer is built for the job and usually beats general upscaling.
- For scanned documents, Upscale PDF works page by page and keeps QR codes scannable.
When upscaling won't help
- The photo is already large enough. It's simply not needed.
- The photo is badly out of focus. AI can't recover what was never captured.
- You need exact fidelity, as in evidence, legal documents or anything where a reconstructed detail could mislead.
Enlarge a small or low-resolution image 2× or 4×, in batches of up to 20.
Open the AI Image Upscaler →Keep reading
- How to restore old family photos
- Resize an image to exact dimensions — free, when you don't need AI