Image Upscaler (2× / 4×)

Standard uses high-quality Lanczos resampling — free, instant, and genuinely good for most enlargements, especially graphics and moderately sized photos. AI reconstructs detail that was never in the original, which makes a visible difference on faces, textures and small or low-resolution photos. It is slower and uses a credit.

Standard mode cannot — it interpolates between existing pixels, enlarging without adding real information. AI mode is different: it predicts plausible detail from patterns learned across millions of images. The result often looks sharper and more convincing, but it is reconstruction rather than recovery of what the camera actually captured.

It differs by mode. Standard handles large files comfortably. AI mode downscales very large inputs to roughly 1024px before processing to stay within GPU limits, so feeding it an enormous image gains nothing. For big photos that only need modest enlargement, Standard is usually the better choice.

Yes, Image Upscaler (2× / 4×) is completely free. No signup, no account, and no watermark on outputs. A Pro tier is available for 100 AI ops/day and larger file sizes.

Image Upscaler (2× / 4×) has two modes, and they send your file to different places. The default mode processes it on our own server and deletes it immediately — no third party is involved. The optional AI mode sends it to our AI provider instead, where it is also deleted after processing and never used for training. You choose the mode before running the tool, so nothing reaches an AI provider unless you pick it.