Bigjpg.AI

Bigjpg.AI splits its upscaling into an Artwork model for illustrations, anime and line art, and a Photo model for camera images. That matters because the failure modes differ: artwork needs hard edges preserved, while photos need grain and compression noise smoothed out. Picking the wrong one shows up fast as either waxy skin or jagged line work.

Paid plans push an image to 16x its original dimensions, with 2x, 4x and 8x in between. A 512px illustration can reach 8192px, which is enough for a large print or a 4K wallpaper. Free accounts are limited to 4x, so 16x upscaling is a paid feature.

The noise slider runs from None through Low, Medium, High and Highest. On a clean digital export, None keeps the original texture. On a JPEG saved five times, the higher settings strip blocking before the network enlarges the frame. Overdoing it flattens fine grain, so the middle settings are the safe default.

Paid accounts can queue many files at once and run several jobs in parallel on high performance servers. Free users process two images at a time on shared hardware and get no batch mode at all. For a shop with 200 product shots, that difference decides whether the work takes an afternoon or a week.

Once you sign in, enlarging happens offline from the browser tab. You can close the laptop, come back later and find the finished file in your history. Anonymous uploads depend on the tab staying open, and the result disappears if you lose the session.

The Bigjpg API exposes four endpoints: request an upload URL, create a task, poll results and retry a failed job. Calls need an X-API-KEY from a paid account, and each response reports remaining_api_calls so you can watch the monthly quota. Style, noise and scale are set per request, which makes it practical to upscale a folder from a script.