AI Image Upscaler
Upscale photos 2× or 4× with ESRGAN — the same AI behind Upscayl. Runs entirely in your browser, no uploads, no account.
Drop your image here
or click to browse · or press Ctrl + V to paste
Frequently asked questions
Does this upscaler upload my image to a server?
No. The ESRGAN AI model runs entirely inside your browser using TensorFlow.js. Your image never leaves your device — all processing is fully local and private.
What is ESRGAN and why is it better than regular upscaling?
ESRGAN (Enhanced Super-Resolution GAN) is a neural network trained to reconstruct realistic fine detail when enlarging images. Unlike bicubic or Lanczos resampling, which just blur the image, ESRGAN hallucinates plausible texture and sharpness — the same model powering the desktop app Upscayl.
What is the difference between 2× and 4× upscale?
2× doubles image dimensions — a 500×500 px image becomes 1000×1000 px. 4× quadruples them. 4× uses a heavier model and takes longer, but produces sharper detail reconstruction, especially useful for very small inputs.
Why is the first run slower?
On first use, the AI model (~15–35 MB) is downloaded and cached in your browser. TensorFlow.js also needs one warm-up inference pass before reaching full speed. All subsequent runs are significantly faster.
What image formats and sizes are supported?
JPEG, PNG, and WebP are accepted. For best results, keep the input under 800 px on the longest side — very large inputs may be slow or trigger memory errors on mid-range devices.