WebTooli

Image Upscaler - Free Online Photo Enlarger with Face Fix

Enlarge photos 2×, 3× or 4× with a proper Lanczos resampler, plus edge-preserving face smoothing that removes pixelation. Runs instantly, no upload, no model download, no dependencies. Free, private, unlimited.

ℹ️ How this tool works. This upscaler uses the Lanczos resampling algorithm in pure JavaScript with zero external dependencies. Before enlarging, an edge-preserving guided filter removes JPEG blocking and noise, mostly on skin (the source of visible pixelation). After enlarging, a skin-only smoothing pass softens just the face and skin tones while keeping eyes, brows, lips and hair sharp, and the rest of the photo gets a stronger detail boost. None of this invents new detail. That requires an AI model, which this lightweight version does not load. There is no server, and your image never leaves your device.
📸
Drop an image here, or click to browse
JPG, PNG, WebP · Max 30 MB · Recommended: under 3000 × 3000 px
Higher factors take longer but produce larger output.
Smooths only the face and skin areas and keeps eyes, lips and hair sharp. The rest of the photo is sharpened, not blurred.
Upscaling…
Upscaled result
Lanczos Resampling — High Quality
💡 At fit-to-screen size, the result looks nearly identical to the original. That is expected. The upscaled image is larger and your browser is scaling it down to fit. Click "Zoom and compare" below to see the actual pixel-level difference.
Original image Upscaled image
Original
Upscaled
Zoom:

Drag either image to move both in any direction. Scroll the mouse wheel over an image (or use + and −) to zoom. Double-click to zoom in. Drag the bar below to resize the box.
Original (browser scaling)
Original photo, zoomed
Upscaled (Lanczos)
Upscaled photo, zoomed
All processing runs locally in your browser using the HTML5 Canvas API. No image is uploaded. No external libraries, no model downloads, no dependencies.

Enlarging a photo usually makes it blurry. Not here.

Most tools stretch an image and call it upscaling. The result is soft. Edges smear. Text becomes unreadable.

This tool uses Lanczos3 resampling. It reads surrounding pixels with a wider kernel than the browser's built-in scaling.

The result is sharper. Text stays readable. Edges stay defined. Texture stays visible.

No server. No upload. No AI model to download. Just math, running in your browser.

Face smoothing that doesn't blur the whole photo

Enlarging also enlarges the flaws. JPEG blocks. Noise. Compression artifacts.

This tool cleans the source before enlarging. Then it smooths only the face and skin tones.

Eyes stay sharp. Eyebrows stay sharp. Lips and hair stay sharp. The background gets a detail boost.

A very light grain is added to skin so it looks natural. Not plastic. Not airbrushed.

You pick the level. Off, Light, Medium, or Strong. If no face is found, smoothing is skipped and the photo is only sharpened.

What this tool does not do

It does not invent new detail. It cannot restore a face that was badly out of focus or extremely low resolution.

That requires an AI super-resolution model. This tool does not include one.

It is a mathematical resampler with edge-preserving smoothing and sharpening. Not a neural network.

If your source is 300 × 300 pixels, a 4× upscale gives you 1200 × 1200 pixels of cleaner, sharper interpretation. Not new information.

How to get the best result

Start with the cleanest source you have. A 1500 × 1500 photo upscales better than a 300 × 300 thumbnail.

Use 2× or 3× for most photos. Use 4× only when the source is already fairly clean.

If the result looks too soft, set Face smoothing to Off or Light. If skin still looks rough, set it to Strong.

Click "Zoom and compare" after upscaling. At fit-to-screen size, the browser shrinks the image and hides the improvement.

The zoom view shows both images at native resolution. That is where the difference is obvious.

Limits and performance

Files up to 30 MB are accepted. The output is capped at 64 megapixels.

If the face area alone exceeds 16 megapixels, skin smoothing is skipped to save memory.

For smooth performance on phones, keep the source under 3000 × 3000 pixels.

A 4× upscale of a 3000 × 3000 photo produces a 12000 × 12000 output. That is 144 megapixels. It will be blocked.

Privacy and speed

Everything runs locally. Your image never leaves your device.

No signup. No account. No watermark. No limit on how many images you process.

No external libraries. No model downloads. Just the browser's Canvas API and JavaScript.

Works on Windows, Mac, Linux, Android, and iPhone. Any modern browser.

Frequently Asked Questions

Why does the upscaled image look the same as the original at first glance?
When an upscaled image is shown in the same container as the original, your browser scales it back down to fit the screen. Click "Zoom and compare" after upscaling to see both images at native resolution. That is where the real improvement is visible.
How does Face smoothing work?
Before enlarging, an edge-preserving filter removes JPEG blocking and noise from the photo. After enlarging, only the face and skin-tone areas are smoothed. Eyes, lips and hair are protected. The rest of the image is sharpened. A very light grain keeps skin looking natural. Choose Light, Medium or Strong. Or Off to disable it.
Will this fix or restore a blurry face?
It can noticeably improve the cleanliness and perceived sharpness of a face using the pixels already present. It cannot restore detail that was never captured, such as a face that is badly out of focus or extremely low resolution. That requires an AI face-restoration model, which this tool does not include.
What is the difference between Lanczos and the browser's default scaling?
Browsers use bilinear interpolation, which averages a 2×2 grid of surrounding pixels. Lanczos uses a larger kernel with sinc-based weighting. That preserves edges, texture and fine lines better than bilinear or bicubic scaling.
Is this a real AI upscaler?
No. It is a Lanczos resampler with edge-preserving smoothing and sharpening. These are filtering and interpolation methods, not a neural network. None of them invent new detail.
Is my photo uploaded or stored anywhere?
No. Everything runs locally in your browser using the HTML5 Canvas API. Your image never leaves your device.
What is the maximum image size I can upscale?
Files up to 30 MB are accepted. The output is limited to 64 megapixels. If the face area alone exceeds 16 megapixels, skin smoothing is skipped to save memory. For smooth performance on phones, keep the source image under 3000 × 3000 pixels.
My photo looks too soft or too rough. What should I change?
If the result looks too soft, choose Light or Off for Face smoothing. If skin still looks rough or pixelated, choose Strong.
← Back to all tools