Image Upscaler
Enlarge and enhance images with AI super-resolution or fast high-quality resampling.
Input is capped to 1024px; larger images are scaled down first, then enhanced.
Higher keeps more of the original photo's texture, countering the smooth/plastic AI look. Lower = cleaner but softer.
Adds subtle grain so skin and surfaces look photographic, not painted.
AI Enhance vs Fast mode
This upscaler offers two engines. AI Enhance runs a Real-ESRGAN super-resolution neural network directly in your browser — it reconstructs edges and texture, so small, soft or low-resolution images come out genuinely sharper, not just bigger. Fast mode uses Lanczos resampling with smart sharpening on our servers; it's instant and handles very large images. Pick AI for quality on smaller images, Fast for speed and big files.
How to upscale an image
Upload a JPG, PNG or WebP image, choose AI Enhance or Fast, set a 2× or 4× factor, and pick an output format. In AI mode a small model downloads once and the whole enhancement runs on your device — your image is never uploaded. A progress bar shows model loading and each processing step, and you can cancel at any time. When it finishes, download the result.
When to upscale images
Upscaling is useful when you have a small logo, product shot, avatar or old photo that looks too small or pixelated for print, a large screen, or a marketing banner. AI Enhance is ideal for rescuing low-resolution or slightly blurry pictures, while Fast mode is best when you simply need a larger version of an already-sharp, high-resolution image.
Restore faces on portraits
For photos of people, AI Enhance has an optional Restore faces mode powered by GFPGAN. It detects each face, rebuilds fine detail like eyes, eyelashes, skin and hair, and seamlessly blends the restored face back into the upscaled picture. Everything runs on your own device — the face model downloads once and is cached — so your photos are never uploaded. It needs a WebGPU-capable browser (Chrome or Edge on desktop) because the model is large.
What an upscaler can and can't do
AI super-resolution recovers a remarkable amount of detail, but no upscaler can invent information that was never captured — a heavily compressed or extremely tiny source still has limits. For the best result, start from the cleanest original you have. AI mode scales inputs down to 1024px before enhancing to keep in-browser processing fast and memory-safe; for very large images use Fast mode, which supports output up to 4000px.
Related Tools
Frequently Asked Questions
What's the difference between AI Enhance and Fast mode?
AI Enhance uses a Real-ESRGAN neural network to actually reconstruct detail and sharpen edges — great for small, blurry or low-resolution images. Fast mode uses high-quality Lanczos resampling plus sharpening — instant, and it handles very large images. AI runs in your browser; Fast runs on our servers.
How does AI upscaling work and is it private?
The first time you use AI Enhance, a small (~5 MB) model downloads and runs entirely inside your browser using WebGPU or WebAssembly. Your image is never uploaded — all the AI processing happens on your own device.
Will upscaling make any blurry photo perfectly sharp?
AI Enhance can recover a surprising amount of detail, but no upscaler can invent information that was never captured. The best results come from small or moderately soft images; a heavily compressed or tiny source still has limits.
Why does AI mode cap the size, and what's the max output?
To keep in-browser processing fast and memory-safe, AI Enhance scales the input down to 1024px first if it's larger, then upscales 2× or 4×. For very large images, use Fast mode, which supports output up to 4000px on the longest side.
The AI result looks too smooth or plastic — how do I fix that?
That's the model's denoising. Use the Detail / natural texture slider to blend the original photo's texture back in, and add a little Film grain — both update the preview instantly without re-running the model, so you can dial in a natural, photographic look.
Balanced vs High detail model — which should I pick?
Balanced (5 MB) is fast and great on faces and small images. High detail (67 MB, RealESRGAN x4plus) resolves sharper hair, beard and fabric across the whole image — pick it for detailed photos and portraits with lots of hair or texture. Both can be combined with Restore faces.
What is the 'Restore faces' option?
In AI Enhance mode you can turn on Restore faces, which uses the GFPGAN model to rebuild eyes, skin and hair on portraits, then blends the restored face back into the upscaled image. It downloads a 340 MB model once (cached afterwards), runs entirely on your device's GPU, and needs a WebGPU browser such as Chrome or Edge on desktop.
Which formats can I upscale and export to?
You can upscale JPG, PNG, and WebP images and export the result as JPG, PNG, or WebP. Choose PNG or WebP to preserve transparency.