About this conversion
Upload an image and open "Details → Upscale" to enlarge it using an on-device AI super-resolution model (a sub-pixel CNN, running as WebAssembly), which fills in detail more convincingly than a plain resize. Everything runs locally in your browser.
Frequently asked questions
How is this different from just resizing to a bigger size?
A plain resize stretches existing pixels and looks soft. This runs a small neural network trained to reconstruct plausible detail when enlarging, which generally looks noticeably sharper.
Is there a size limit?
Very large images take longer and use more memory since the model runs on your own device’s CPU/GPU rather than a server. For huge source images, expect it to take a few seconds.
Is my image uploaded to run the AI model?
No. The model (ONNX Runtime Web) runs entirely on-device via WebAssembly. Nothing about the image is transmitted.