Question 1

How does the upscaling work?

Accepted Answer

The tool uses a deep learning super-resolution model (ESRGAN). Instead of simple stretching, a neural network infers and restores fine details and textures for the upscaled image, resulting in sharper edges and clearer visuals. The model runs in your browser using TensorFlow.js, making it suitable for upscaling photos, illustrations, and restoring old images.

Question 2

Why is it slow the first time and requires internet?

Accepted Answer

The first time you use it, you need to download the AI model files (tens of MB) from the CDN. Once downloaded, the browser caches them, making subsequent uses much faster. Because the model is large and inference requires computing power, processing large images may take time. We recommend uploading moderately sized images. You can also start with a 2x upscale to check the result before deciding whether to use 4x.

Question 3

Will my images be uploaded to a server?

Accepted Answer

No. Only the AI model files are downloaded. Your images are always processed locally in your browser and are never uploaded to any server. This is fundamentally different from other upscaling services that require uploading images to the cloud, offering better privacy protection.

Lossless Image Upscaler

FAQ