AI image upscaler

What AI upscaling does

Normal enlarging stretches the pixels, so the result looks soft and blocky. An AI upscaler predicts the missing detail. This page uses Real-ESRGAN, a model trained to rebuild edges, remove JPG compression artifacts and reduce noise while it enlarges. It works well for small web images, old phone photos, screenshots, game art and scanned drawings.

The model runs in your browser. Your images are not uploaded.

How to upscale an image

1. Drop one or more images in the area above, or paste one with Ctrl + V. The scale is set to 200% with the AI method.
2. Choose AI photo for photos, or AI illustration for drawings, anime, comics and logos.
3. Click 4× or move the scale slider if you need a larger result. The slider goes up to 400%.
4. Drag the divider in the preview to compare the original and the result. Use View actual size to check fine detail.
5. Click Save for the current image, or Save all for every image with the same settings.

Speed and requirements

The model downloads once on first use (about 5 MB). With GPU access through WebGPU, most images finish in seconds. On browsers or devices without WebGPU, the AI step can be slow. If it cannot run at all, the tool tells you and uses the basic method instead. Recent versions of Chrome and Edge on a desktop give the best results.

Tips for good results

• Start from the largest version of the image you have. AI fills in detail, but it cannot recover text or faces that are only a few pixels wide.
• Save as PNG to keep the restored detail. Save as JPG at quality 90 or higher if the file has to be small.
• The AI method applies only when the image gets larger. At 100% or below, the tool resizes normally.
• Output is capped at 12,000 px per side and 100 million pixels in total, because of browser limits.

For print, work out the size you need first. A 6×4 inch print at 300 DPI needs 1800×1200 px, so a 900×600 image needs 200%. Upscaling past the size you need only makes the file larger.