Image Enhancer

AI upscale, sharpen, denoise and fix color.

Your image is processed on this device and is never uploaded.

About this tool

Load a photo and enhance it with presets or fine-grained sliders, comparing before and after live. Noise is removed first with an edge-preserving guided filter, tone and color are corrected (auto levels, gray-world white balance, exposure in linear light), the image is upscaled with Lanczos-3 resampling, and clarity and sharpening are applied last at the final resolution. For small, blurry or heavily compressed photos, turn on AI super-resolution: a Real-ESRGAN neural network runs on your GPU (WebGPU) or CPU and rebuilds sharp edges instead of just enlarging pixels. Everything runs in background Web Workers on your device - nothing is uploaded.

How to use

  1. Drop, browse for, or paste (Ctrl + V) an image.
  2. Pick a preset such as Auto enhance or Old photo, or tune the sliders by hand.
  3. Choose an upscale factor, and turn on AI super-resolution for small or blurry photos (raise Artifact cleanup if the result looks blocky).
  4. Drag across the preview to compare before and after.
  5. Pick PNG, JPEG or WebP and download the full-resolution result.

Examples

Pipeline
denoise → tone & color → upscale → clarity → sharpen
Upscale
1200×800 → 2400×1600 (2x, Lanczos-3)
AI mode
cleanup (deblock + soften) → Real-ESRGAN 4x → resample to target → your adjustments

Frequently asked questions

No. Your image is decoded, processed and re-encoded entirely by your browser using a background Web Worker. Nothing is sent to a server.
The regular controls can't: they reveal, clean up and sharpen the detail that's already there. AI super-resolution can: its neural network predicts plausible edges and texture, so small or blurry photos come out much crisper. Predicted detail isn't recovered truth, though - very damaged images can pick up a painted look, and it won't restore faces or text that are unreadable in the original.
It runs Real-ESRGAN (realesr-general-x4v3, BSD-3-Clause), a compact network trained on real-world blur, noise and JPEG damage, with ONNX Runtime Web. It uses your GPU through WebGPU when the browser supports it (current Chrome and Edge) and falls back to the CPU, which is much slower. The model (about 5 MB) and engine (about 27 MB) are downloaded from this site once and cached; your image never leaves the device. AI mode handles photos up to 16 MP with WebGPU, or 1 MP on the CPU; at 1x it restores the photo at its original size.
For large images, the live preview is computed at up to 1200 pixels on the long side so the sliders stay responsive. The download is always processed at full resolution, so fine sharpening looks a little crisper in the saved file.
Source images up to 50 megapixels, and outputs up to 36 megapixels. Upscale factors that would exceed that are disabled. Very large exports need a lot of memory, and some mobile browsers have lower canvas limits.
PNG is lossless and best for further editing or graphics. JPEG and WebP make much smaller files for photos; a quality around 90 is visually lossless for most images. JPEG has no transparency, so transparent areas become white.
Contact

Let’s connect

Have a Liferay or Backend project you’d like to talk through? Fill in the form and I’ll get back to you.

Book Calendar