How to increase image resolution without losing quality

A digital image is a grid of pixels. A 1200 × 800 image has 1200 pixels across and 800 down, about 1 megapixel. That pixel count is its resolution, and it decides how large the image can be shown or printed before it looks soft.

The DPI or PPI number stored in a file is something else. It only tells a printer how many pixels to place on each inch of paper. Changing it from 72 to 300 makes the planned print smaller; it adds no pixels and no detail. The 300 DPI guide explains this in full.

Every editor can make an image bigger: Image Size in Photoshop, Resize in Windows Photos, Preview on a Mac, or any online resizer. They all use interpolation. The program spreads the existing pixels over a larger grid and fills the gaps with averages of their neighbours.

The new pixels contain no new information, so the result is bigger but evenly soft. Edges turn into gradients, small details melt together, and any compression blocks in the original are enlarged with the picture. Doubling the size this way is often acceptable; going to four times looks clearly blurred.

An AI upscaler is a neural network trained on pairs of sharp photos and small copies of them. It learns what a sharp edge, a strand of hair or a patch of grass usually looks like at a larger size, and draws the enlarged image to match the shapes in your picture.

Upscale 4× makes an image four times wider and taller, so it has sixteen times the pixels: 1000 × 750 becomes 4000 × 3000. Upscale 8× goes eight times wider and taller for very small sources such as thumbnails. Both keep the composition, colours and faces of the original rather than redrawing them.

The fine texture the model adds is plausible, not recorded. A brick wall gets brick-like texture that follows the bricks in the photo, but it is not the exact texture the camera would have captured at that size.

Work out the target first, then choose the smallest enlargement that reaches it. Some common targets:

• Full HD screen or slide: 1920 × 1080. 4K screen: 3840 × 2160 (see how to upscale an image to 4K).
• Photo print at 300 DPI: the print size in inches times 300. A 6 × 4 print needs 1800 × 1200; an 8 × 10 needs 2400 × 3000. The print size chart lists every common size.
• Poster seen from a distance: 150 DPI is usually enough, so 24 × 36 inches needs 3600 × 5400 (poster sizes in pixels).
• Product photos with zoom: aim for at least 2000 pixels on the shortest side (product photo guide).

Find the largest copy of the image you can. The original from a camera roll or the first download beats a version saved from a chat, a screenshot or a social media post, which have already been shrunk and recompressed.

Look at it at 100 % zoom. If you see grain or square compression blocks, run Enhance first: it cleans them at the same size, and the upscaler then enlarges clean detail instead of damage.

Divide the target size by the image size. If the factor is four or less, use Upscale 4× (images up to 5 MP). If it is between four and eight, use Upscale 8× (images up to 1.25 MP, 2 credits). Crop or reduce the result to the exact size you need.

Check faces, edges and any text at 100 % before you print or publish, and save as PNG or a high-quality JPEG so you do not add new compression.

No tool, AI or not, recovers information that is completely missing. A face that is ten pixels wide does not become a recognisable portrait; we do not invent one, because it would not be the person in the photo. Text that is already unreadable stays unreadable. Strong motion blur or a missed focus stays blurred; see can a blurry image be unblurred.

Running an upscaler again on its own result adds pixels but not real detail. If one 8× upscale does not reach your target, choose a smaller print or find a larger original.

You get 3 free credits a day, with no sign-up and no card. One photo with one tool uses 1 credit; Upscale 8× uses 2. Free uploads can be JPG, PNG or WebP up to 3 MB and 8 MP, and free results carry a watermark. Pro ($9.99 a month, 200 credits) and Pro+ ($19.99 a month, 500 credits) remove the watermark and accept uploads up to 20 MB and 30 MP. Uploads and results are deleted within 24 hours and are never used to train AI.