Image Upscaler
Runs a Swin2SR super-resolution model in your browser to enlarge the image 2× or 4× with detail plain scaling can't recover.
What this does
Image Upscaler runs the Swin2SR super-resolution model directly in your browser through Transformers.js — your image is never uploaded to a server.
The first time you click Upscale, the model downloads from the Hugging Face CDN and is cached (2× and 4× use separate models). Upscaling runs in a background worker, so the page stays responsive. Links to the model and library are below.
How it works
- 1Drop an image.
- 2Pick 2× or 4×, then click Upscale. The first run downloads the super-resolution model once and caches it.
- 3Download the enlarged PNG.
Built with open source
- Transformers.js — Hugging Face's library for running ONNX machine-learning models in the browser, on WebGPU or WebAssembly. The model weights download from the Hugging Face CDN on first use and are cached. · Apache-2.0
- Swin2SR — Super-resolution transformer that enlarges images while reconstructing detail. · Apache-2.0
Frequently asked questions
Related tools
All Edit image →Remove BackgroundRemove the background from an image and download a transparent PNG. Runs entirely in your browser.Resize ImageResize an image to any width and height. Lock the aspect ratio so nothing stretches.Crop ImageCrop an image by dragging a box around the exact part you want to keep.Blur BackgroundBlur the background of an image while keeping the subject sharp. Runs entirely in your browser.Pixelate ImagePixelate an image by averaging it into square blocks. You set the block size.Grayscale ImageConvert an image to grayscale, mapping each pixel to its luminance.
