Guides / Upscaling

Image upscaler vs AI upscaler

Search results mix two different products. A classic image upscaler resamples pixels you already have. A generative AI upscaler predicts new texture. PixTools ships the first kind, on purpose.

What “upscale” used to mean

Before diffusion models, enlarging a photo meant interpolation: bicubic, Lanczos, or similar filters, often followed by an unsharp mask. The file gets more pixels. Edges can look crisper. Soft blur in the original stays soft — just bigger. That is what PixTools does at 2x and 4x, entirely in the browser, with no API key.

What generative AI upscalers do

Paid “AI upscale” tools run a neural network that was trained to invent plausible detail: pores, stitching, foliage. Results can look stunning on portraits and disastrous on text, logos, or product labels (hallucinated letters). Those services usually upload your file to a GPU cluster. That can be the right trade if you need invented texture and you trust the vendor.

Side-by-side, without marketing fog

When to use PixTools anyway

Use the in-browser enlarge if you need a 2x/4x file quickly, you do not want the original on someone else’s disk, or the image already has the detail you care about (UI captures, QR-adjacent graphics, a product shot that is merely too small for a slide). Crop first with the image cropper if you only need part of the frame.

People searching “free AI image upscaler” often want the generative kind. We keep that phrase in the FAQ so the difference is explicit, not so we can pose as a diffusion demo.