Image Upscaler: Enlarge images 2× or 4× with sharp resizing or AI enhancement. Input: Image files and values you enter. Output: Results shown on the page and a downloaded file. Privacy: Runs in your browser: inputs and files stay on your device and are not uploaded.
How to upscale an image
- Drop a photo, logo, screenshot or illustration onto the box.
- Choose a method. Sharp resize enlarges any image instantly with high-quality resampling and adjustable sharpening. AI enhance uses the Real-ESRGAN neural network to rebuild fine detail and remove blur and JPEG blockiness; it is slower and works on images up to about 1.2 megapixels with a graphics card (WebGPU), and smaller ones elsewhere.
- Pick 2× or 4× and press Upscale. For AI enhance, the model downloads once (the size is shown first) and the image is processed tile by tile with a progress bar.
- Drag the before/after slider to compare, then download as PNG or JPG.
What upscaling can and cannot do
Enlarging an image means inventing pixels that were never captured. Ordinary resizing averages neighbouring pixels, which keeps the picture faithful but soft; sharpening then restores some edge contrast. An AI upscaler such as Real-ESRGAN has learned from millions of photos what sharp detail usually looks like, so it can turn a small, blurry or heavily compressed image into a cleaner, crisper one. It is guessing, though: it cannot recover a face, text or a licence plate that is unreadable in the original, and it can make textures look slightly painted.
Good uses are enlarging old photos and small web images for printing, making product photos meet marketplace minimum sizes, cleaning up low-resolution logos and screenshots, and preparing images for slides and posters. Both methods run entirely in your browser, so your images stay private, with no watermark or credits. AI enhance is much faster on computers with a modern graphics card (WebGPU); on phones, prefer sharp resize or small images.
Is my image uploaded?
No. Both methods run on your device. AI enhance downloads the open Real-ESRGAN model (BSD licence, about 17 MB) once from this site.
Why is AI enhance limited to smaller images?
The neural network needs a lot of computing power per pixel; 1.2 megapixels in already becomes 19 megapixels out at 4×. Without a graphics card it runs on the processor, which is much slower, so the limit there is about 0.25 megapixels. Crop to the important part for bigger photos.
Which should I choose, 2× or 4×?
2× for most uses. 4× suits very small images, such as icons or thumbnails, that need to become print-sized.
Can it make a blurry photo sharp?
It can make it look sharper and cleaner, but it cannot restore details that are completely missing.