Drop an image here
or click to choose · PNG, JPG, WebP, GIF
Choose imageAbout the Image Upscaler
Enlarge a picture without the mush: this upscaler uses an ESRGAN super-resolution neural network to grow images to double or quadruple size, reconstructing edges and texture instead of just stretching pixels the way ordinary resizing does. Pick 2× or 4×, check the projected output dimensions shown above the preview, press Upscale, and follow the progress bar as the AI works through your image patch by patch.
It's built for rescuing small images — old photos, tiny logos, game assets, thumbnails saved without originals — and for preparing web-size pictures for print. Like Tooldoodle's background remover, the neural network runs inside your browser: the model downloads once on first use, is cached afterwards, and your image is never sent to a server. Free, no sign-up, and the enlarged result saves as a PNG.
Features
- ESRGAN neural network for genuine detail reconstruction
- Choose 2× or 4× enlargement to suit the source
- Output dimensions previewed before you run it
- Patch-by-patch processing with a live progress bar
- AI runs in your browser — images stay on your device
- Free unlimited upscaling with PNG download
How to upscale an image with AI online
- Drop the image you want to enlarge onto the tool.
- Choose 2× or 4× and check the projected output size.
- Click Upscale — the model downloads on your first run.
- Watch the progress bar as the AI processes the image.
- Compare the result, then download the enlarged PNG.
Frequently asked questions
How is AI upscaling better than normal resizing?
Conventional resizing interpolates between existing pixels, which inevitably blurs edges as the image grows. The ESRGAN model was trained on millions of image pairs to predict what the high-resolution version should look like, so it reconstructs plausible edges, textures and fine detail rather than smearing what's there.
Should I pick 2x or 4x?
Use 2× for moderate enlargement and the most natural results — it's also faster. Reserve 4× for genuinely tiny sources like thumbnails and small logos, where a 300-pixel image becomes 1200 pixels. On already-large photos, 4× produces enormous files and processing times without much visible benefit.
Why does the first upscale take longer?
Before anything can run, the neural network weights are downloaded into your browser and initialised — that's the one-time wait the status message describes. The model is cached afterwards, so future sessions jump straight to processing, whose speed then depends on your image size and hardware.
Is there a maximum image size for upscaling?
No fixed cap, but memory and time are real constraints since everything runs on your device: the image is processed in small patches, and a large photo at 4× can mean thousands of them. The tool suits small-to-medium images best; a multi-megapixel photo rarely needs AI enlargement anyway.
Are my photos uploaded for the AI to process?
No. The super-resolution model executes in your browser using your device's own processor and graphics hardware — the only download is the model itself, and the only thing leaving the page is your finished PNG when you save it. Old family photos stay as private as they were in the shoebox.