Image target file size.

Aim for a byte budget. See the actual result.

Files stay on your device

Choose image

Drag and drop one image here, or

JPEG, PNG, WebP and other browser-decodable images

No image selected.

Target settings

Largest output you will accept. Sizes on this page use 1 KB = 1,000 bytes; pick KiB or MiB if your limit counts 1,024. Fractional bytes round down.

This target is 200,000 bytes.

The search looks for the highest whole quality point in this range that fits. Lower settings lose more detail; even 100 can change pixels.

Original dimensions are kept unless you enable resizing.

Choose an image to begin.

Export creates one still image in 8-bit sRGB. Source metadata is removed; HDR and wide-gamut colors can change. Check the previews before downloading.

Compare the result

Preview both images, then choose what to keep.

Original decoded still frame

No image selected.

Result decoded output

Your result will appear here.

Actual target status and byte savings will appear after encoding.

No automatic downscaling. Dimensions change only when you enable resizing. PNG and WebP retain transparency; JPEG uses your selected background.

Preview images fit these panels; downloads use the reported dimensions. Each new run starts from the original decoded image.

How it works

  1. Add your image. Check the decoded still frame and original size.
  2. Set your target. Choose format and quality range. Enable resizing only if you want fewer pixels.
  3. Check and download. Compare actual bytes, dimensions and preview, whether or not the target was achieved.

What gets tested

JPEG and WebP are encoded at your maximum quality first. If that is too large, your minimum is tried, and the search then halves the range between them until it finds the highest tested quality at or below your target, usually in under ten encodes. If even the minimum is too large, you get the smaller of the tested results with “Target not achieved.”

Every offered file is measured, never estimated. The search assumes size grows with quality; encoders occasionally break that, so an untested quality point could also fit. This is not an exact byte count or a guarantee of visual quality. Dimensions stay fixed throughout the search. PNG encodes once without a quality setting.

Files stay on your device

Decoding, resizing and encoding run in your browser. This tool does not upload, log or store image contents or filenames.

Cancel stops the search and keeps your image and settings. Reset releases the image and restores default settings. New settings discard an outdated result; every search starts from the original decoded pixels.

Frequently asked questions

What if my target cannot be reached?

The tool reports the measured output bytes and target separately. “Target not achieved” means none of the tested encodes fits; it does not prove every possible encoder would fail. Try a lower minimum quality, a different format, or explicitly enable a smaller longest edge. Small images and already optimized files may become larger; the result says when the output is larger than the source, and when your original already fits the target.

Will my image be resized or lose transparency?

No resizing occurs unless you enable it. The selected longest edge downsizes proportionally and never enlarges smaller images. PNG and WebP retain transparency. JPEG replaces transparent areas with the selected background color. PNG losslessly encodes the decoded canvas pixels, but resizing and color conversion can still change the source.

What happens to metadata, color and animation?

Orientation is applied before export. Animated and multi-image inputs become the single still frame shown in the original preview. Source EXIF, GPS and other metadata are not copied. Output uses 8-bit sRGB canvas pixels, so HDR, wide-gamut and high-bit-depth detail can change.

Which browsers and file sizes work?

Use a current browser for the widest format support. AVIF and images your browser can open are supported regardless of their filename. Large images can exceed browser memory or canvas capabilities; errors are reported without silently shrinking the image. There is no tool-imposed byte, pixel or time cap.