Guide
How to upscale images with local AI
AllPic enlarges images 2x or 4x with Real-ESRGAN detail reconstruction running entirely in your browser. It works for photos, screenshots, product shots, and illustrations - and because the model runs on your device, nothing is uploaded, and there is no sign-up or watermark.
Recommended flow
- Step 1
Open the AI Image Upscaler and add a photo, screenshot, or illustration.
- Step 2
Pick 2x for sharing and print, or 4x for heavy crops and old photos.
- Step 3
Run the model locally - the first run prepares about 5 MB once.
- Step 4
Compare before and after with the slider, then download.
When to choose 2x
2x is the safe default for chat covers, marketplace photos, and slide decks. It doubles the resolution, keeps memory use low, and finishes fastest - especially with WebGPU.
When to choose 4x
4x suits small source images: old photos, thumbnails, logos saved too small, or heavy crops. Combine it with AI Auto Enhance for old photo restoration.
Screenshots and UI images
The general model keeps text edges readable. For documentation images, upscale 2x and export PNG so compression does not blur the reconstructed text.
Export format tips
JPG keeps 4x results small for sharing. PNG preserves every reconstructed pixel and transparency. WebP is a good middle ground for the web.
FAQ
Useful limits and answers
Is the image uploaded to a server?
No. The Real-ESRGAN model runs inside your browser with WebGPU or WebAssembly, so the photo never leaves your device.
What is the difference from normal resizing?
Normal resizing stretches existing pixels and looks blurry. The AI model reconstructs edges and texture, so text, faces, and product edges stay crisp.
How do I make an old photo clearer?
Upscale it 2x or 4x first, then run AI Auto Enhance to recover contrast and color. Both steps run locally.
Why was my very large image reduced first?
The tool keeps the output within a safe browser memory budget. It tells you when the input was reduced before upscaling.
Does it need a special browser?
Any modern browser works. Chrome and Edge with WebGPU are fastest; other browsers use the WebAssembly path.