Compress a JPG to 50KB

Staying in JPG is the point of this page. Better formats exist, but the places that impose a fifty kilobyte ceiling — application portals, banking forms, older content systems — routinely accept nothing else, so the job is to get the most out of a format designed in the early nineties.

How should the settings be chosen?

    What JPG does badly, and how to avoid it

    JPG works on eight by eight blocks of pixels and describes each block as a set of frequencies. When the budget runs out it keeps fewer frequencies, and the blocks start to become visible as squares, particularly across smooth gradients like a sky or a plain wall behind a portrait.

    It also struggles with sharp edges, which is why text and line art acquire a faint halo. If your image is mostly text, JPG is the wrong tool regardless of the size limit and PNG will be both smaller and cleaner.

    Give the encoder less to do

    Reducing the pixel width is far kinder to a JPG than reducing quality, because it removes blocks rather than degrading each one. At a fifty kilobyte target, eight hundred pixels across is a reasonable ceiling for a portrait and leaves the quality search working in a comfortable range.

    Cropping helps for the same reason. Everything you remove from the frame is budget returned to the part of the picture you care about.

    Re-compressing an existing JPG

    A JPG that has already been compressed is not a clean starting point — the block artefacts from the first pass are now part of the picture, and the second pass has to encode them faithfully. Work from the largest, least processed version you have. Where the only copy is already small, check the result against the original before replacing it.

    Questions

    Why does my JPG look blocky after compressing to 50KB?
    Because the encoder ran out of budget and started discarding detail within each eight by eight block, which then become visible as squares. Reduce the pixel width instead of pushing quality lower — fewer blocks at a decent quality looks considerably better than many blocks at a poor one.
    Does compressing a JPG twice make it worse?
    Yes. Each pass discards information permanently, and the second pass must also encode the artefacts introduced by the first, which wastes budget on flaws. Always start from the least processed copy you have rather than from a file that has already been reduced once.
    Can I convert to WebP instead to get better quality at 50KB?
    You would get a visibly better picture, because WebP is a far more capable format at that budget. Whether you can depends on the destination: many upload forms accept JPG only and reject anything else. Where the destination is a website or a chat app, WebP is the better choice.