Upscale images with AI
Enlarges images 2× or 4× and reconstructs detail instead of just enlarging pixels.
or tap to choose. Best for small or soft images.
Choose fileSettings
Embed this tool for free
Paste this code where the tool should appear on your website. It contains your personal key.
Free, no sign-up. The tool runs in your visitors' browsers and their files are never uploaded. A small credit to enhancevision.de appears below the tool.
Step by step
- Drop the image. It appears in the preview right away, and below the settings you see its current size and the size after upscaling.
- Choose 2× or 4× under “Upscale factor” and PNG or JPG under “Format”.
- Tap “Upscale”. The image is processed tile by tile; the bar shows progress in percent.
- Compare the result using the “After” and “Before” tabs. The before view deliberately shows the original unsmoothed at the result size so the difference is easy to see.
- Save with “Download”. The file name gets -2x or -4x added. With “Continue with …” you can open the result straight in another tool.
What you can use it for
- Photo books and prints: A phone photo from a chat or a crop is fine on screen but looks rough in print. Upscaled, edges and surfaces look calmer in larger formats.
- Enlarge AI images: Many image generators output fairly low resolutions. With 4× you get a version better suited to wallpapers, posters or further editing.
- Small product and archive images: Old shop images or photos from earlier websites often only exist at thumbnail size. Upscaled, they can be used in current layouts again.
- First step for other tools: With “Continue with …” the enlarged image goes straight to background removal, cropping, converting or laser engraving prep, without saving it in between.
When does upscaling help?
- Images too small to print: such as a web image you want on canvas or as a poster.
- Laser and plotter templates: more pixels give cleaner edges on large engravings.
- Old photos from the web: the AI smooths out compression artifacts.
The AI does not invent real detail that isn't in the image. Faces and text from very small originals therefore stay limited.
Frequently asked questions
How large can the original be?
That depends on your device. The image is processed in tiles, so larger images work too, but take longer.
Is my image uploaded?
No. The AI model runs in your browser.
Which model is used?
Real-ESRGAN (realesr-general-x4v3), a free model for general photos.
Is 2× faster than 4×?
No. The model always computes 4×; for 2× the result is then cleanly halved. Processing time is the same, 2× just gives a smaller file.
What happens to transparent areas?
Transparency isn't kept; see-through areas are filled with white before upscaling. If you need a cut-out subject, upscale first and remove the background afterwards.
PNG or JPG?
PNG is lossless but the file gets large. JPG is saved at high quality and is much smaller, usually the better choice for photos and sending.
Why does it say the image will be reduced first?
If the result would be too large for the browser's memory, the original is reduced accordingly before upscaling. The sizes shown below the settings already take this into account.
Why does the status line sometimes say WebGPU?
Then your browser uses the graphics card for the calculation, which is usually much faster. Without WebGPU the processor does the work; the result is the same.