HomeGuides › Upscale Image

How to upscale an image with AI

Enlarge small photos, screenshots, and artwork with a neural network that reconstructs detail — up to 8K, without your image leaving your device.

Plain enlargement stretches pixels: a small image gets bigger and blurrier. AI super-resolution is different — a neural network trained on millions of photos predicts the detail that should be there, producing sharper edges and cleaner texture. Toolsfully runs this AI directly in your browser, which is rare: your image is never uploaded to anyone's server, even for the heavy lifting.

Step-by-step

  1. Open the Upscale tool. Go to Image Tools and select the Upscale tab.
  2. Add your image. Small and medium images benefit most — that's where the AI has detail to reconstruct.
  3. Pick a method and scale. AI super-resolution gives the best quality at 2×, 3×, or 4×. Instant enlarge skips the AI for a fast plain resize when you just need bigger dimensions.
  4. Upscale and download. The first AI run downloads a ~3 MB engine, then processing runs on your device with live progress. The result saves as a PNG.
AI processing speed depends on your hardware — a small photo takes seconds on a modern laptop, while 4× on a large image can take a few minutes. The progress percentage keeps you posted, and output is capped at 8K (7680 × 4320).

Tips

Frequently asked questions

Is it really AI, and really local?

Yes — an ESRGAN-family super-resolution network runs via TensorFlow.js inside your browser. You can watch your network tab: after the engine loads, nothing is transmitted.

Why is my output capped below 8K?

The tool limits output to 7680 × 4320. If your image × scale would exceed that, it's automatically reduced to fit — the status line tells you when that happens.

Why does it take so long on my machine?

Inference runs on your device's GPU/CPU. Older hardware and bigger images mean longer runs — try 2× or the instant mode.

Open the Upscale tool →