Compress JPG

Lower quality usually makes a smaller file. A maximum width above zero downsizes wider images without enlarging smaller ones.

Re-encode up to 10 genuine JPEG files in your browser, choose quality from 10 to 100, and optionally reduce images wider than your chosen limit. The result card reports the exact original and output byte sizes, including when re-encoding makes a file larger.

How to compress a JPG

  1. 1

    Choose the settings

    Set JPEG quality from 10 to 100 and, if needed, a maximum output width from 1 to 12,000 pixels. Zero keeps the decoded width.

  2. 2

    Choose the JPEG files

    Select up to 10 .jpg or .jpeg files, 20 MB each and 100 MB in total. Signature, dimensions and browser decoding are checked before processing.

  3. 3

    Review each result

    Inspect the compressed preview, output dimensions and exact size change. A result can be larger than its source.

  4. 4

    Download new JPG files

    Download one result or all results. The original files are never overwritten or uploaded.

What the quality setting means

JPEG quality is an encoder hint, not a promised percentage or a standard shared by every browser. A lower value usually removes more detail and produces a smaller file, but image content and the source encoder matter. Quality 100 is still a lossy re-encode. Judge fine text, hair, skies and hard edges in the preview rather than relying on one universal setting.

Why a compressed file can be larger

An existing JPEG may already be highly optimised. Re-encoding it at a high quality, or keeping all its pixels, can produce more bytes than the source. The tool reports that increase honestly. Lower the quality or set a smaller maximum width if a smaller file is the priority.

Limits and output

Each source is limited to 20 MB, 12,000 pixels on either side and 40 megapixels. A batch can contain up to 10 files totalling 100 MB. Files are processed sequentially, and each output must be a valid JPEG no larger than 25 MB. The output always uses the .jpg extension and image/jpeg media type.

Orientation, metadata and privacy

Common camera rotations in EXIF orientations 5 to 8 are normalised into the output pixels after decoded dimensions are verified. Square images with tags 5 to 8 are rejected because decoder rotation is dimensionally ambiguous; mirrored or 180-degree tags 2 to 4 are rejected too. Canvas re-encoding removes EXIF, GPS, comments and embedded colour profiles. Each result is limited to 25 MB and the output batch to 100 MB. Files never leave the browser. In the step-by-step flow, results are restorable and usable for 30 minutes. An active page deletes them at expiry; if the tab was closed or suspended, the next funnel visit lazily deletes the expired record.

Avoid generational loss

Every time you save a JPEG you lose a little more detail. Never re-compress the same JPEG repeatedly, each round compounds the artifacts. Edit from the original RAW/PNG/TIFF whenever possible and treat the JPEG as a final export only.

When lower quality is actually fine

  • Thumbnails and list previews, nobody zooms in.
  • Social media uploads, platforms re-compress anyway, so shipping a 95-quality file is wasted bytes.
  • Background images, viewers are not scrutinising them.

Frequently Asked Questions

Start at the tool’s default of 80, inspect the preview, then adjust for your image and destination. There is no universally lossless-looking value, and browsers may encode the same setting differently.

Yes. This tool decodes and re-encodes the pixels, so quality 100 is still lossy and can also create a larger file. Keep an original or use a lossless workflow when every source pixel matters.

Yes. Canvas output does not copy EXIF, GPS, comments or embedded colour profiles. The files are processed locally and are not uploaded. Funnel results in IndexedDB are restorable for 30 minutes; an active page deletes them at expiry, while a closed or suspended tab leaves cleanup to the next funnel visit. Check the preview because stripping a colour profile can affect appearance.

Consider them when your destination accepts them, but compare the actual outputs. Quality scales are not directly equivalent across JPEG, WebP and AVIF, so “the same quality” cannot be inferred from the same slider number.

Related Tools