Enlarge and sharpen a photo 2× or 4× with on-device AI super-resolution. It runs the open-source Swin2SR model in your browser with WebGPU, so nothing is uploaded. Free, no account, no watermark.
No account · no install · nothing uploaded · MIT open source
# upscale a photo in 4 steps 1. Drop an image png · jpg · webp 2. Pick a model ×2 fast · ×4 high 3. Run on-device · WebGPU 4. Download the upscaled image
Opens the 🖼️ Image tab → Image Upscaler. First run downloads the model, then it's cached.
From a small or soft photo to a sharper, larger one — in four steps, with no upload.
Choose or drag a PNG, JPG or WebP. It loads into a private in-browser workspace — your file never leaves the device.
×2 (Swin2SR) is fast for general photos; ×4 is higher quality and restores compressed / JPEG images. Downloads once, then cached.
Super-resolution runs with WebGPU (or WebAssembly), reconstructing detail as it enlarges — right in the tab.
Compare before/after with the wipe slider, then download the upscaled image.
Open the tool, drop an image, choose a 2× or 4× model and click Run. The AI enlarges and sharpens it and you download the result — free, no account, in your browser.
No. The super-resolution model runs on your device with transformers.js (WebGPU, falling back to WebAssembly). Only the public model weights are downloaded once and cached; your image stays local.
2× with the fast Swin2SR model, or 4× with the higher-quality model that also restores compressed / JPEG images.
Super-resolution is compute-heavy and runs on your own device, so very large images take longer — especially without WebGPU. Try the 2× model or a smaller input first.
Neither — slicerflow is free and open source, with no watermark and no account.
On-device AI super-resolution — free and private, in your browser.