Upscale Images With AI in Your Browser
Make a small image 2, 3 or 4 times bigger without the blocky, blurry look of simple stretching. AI mode runs an ESRGAN neural network on your device that adds convincing detail to photos, text and textures; Smooth mode uses high-quality stepped resampling and works on large images. Either way, the image stays in your browser.
Your image is processed by your own browser and never uploaded. Many online editors send your photo to their servers and delete it later — this one never receives it at all.
Why use Image Upscaler
- AI mode: an ESRGAN neural network running in your browser
- Smooth mode: stepped high-quality resampling with optional sharpening
- 2x, 3x and 4x enlargement
- Checks the output size against your device's limits before starting
- Save as the original format, JPG, PNG, WebP or AVIF
- Nothing is uploaded
Popular uses
- Enlarging a small logo or icon for a presentation
- Making a low-resolution product photo usable in a larger layout
- Scaling up screenshots for documentation
- Preparing a small image for printing
Image Upscaler: frequently asked questions
How does the AI upscaler work?
AI mode uses ESRGAN, a neural network trained to rebuild fine detail when an image is enlarged. It runs in your browser with TensorFlow.js, so edges, text and textures come out sharper than with resampling alone.
What's the difference between AI and Smooth mode?
AI mode adds detail and gives the sharpest result, but it's slower and works on images up to 1.5 megapixels. Smooth mode enlarges images of any size quickly with high-quality resampling and optional sharpening, without inventing new detail.
Why is AI mode limited to 1.5 megapixels?
The network runs on your own device, and bigger images would take minutes and a lot of memory. For larger photos, use Smooth mode, or crop to the part you need first.
Can AI upscaling recover details that aren't in the photo?
Not truly. The network predicts plausible detail from patterns it learned, which looks sharper, but it can't restore information that was never captured, such as text that's unreadable in the original.
Is my image uploaded?
No. Both modes run in your browser, and the AI model is downloaded to your device once.