Target a byte budget · Compare before and after · Local processing

Image Compressor for a Target File Size

Compress one static JPEG, PNG, or WebP or queue a local batch of up to 20 files. Keep the highest JPEG or WebP quality found under the target, then download individual results or one ZIP.

Compress images20 KB to 1 MB targets · Up to 20 files · No sign-up
measured result
input   2,048 × 1,365 · 1.8 MiB
target  ≤ 200 KB
output  reported after encoding

Target, not exact sizeQuality and dimensions stay visible

Compress one image or a local batch of up to 20

Find the best result the selected target permits

Choose validated static images, a target such as 100 KB or 1 MB, and an output format. Preview each encoded result and its measured byte size before downloading it.

Loading the image compressor…

Enable JavaScript to select, validate, encode, preview, and download an image. The target-size method and format limitations remain readable below.

Measure every candidate

How target-size image compression works

A file-size goal is an optimization problem, not a metadata label. The output must actually be encoded before its bytes can be counted.

  1. 1

    Validate and decode

    The tool checks PNG, JPEG, or WebP signature bytes instead of trusting a renamed extension. It then decodes the static image and rejects unsupported, animated, oversized, or undecodable input.

  2. 2

    Search quality where it exists

    For JPEG and WebP output, candidates are encoded at different quality values. The compressor retains the highest-quality candidate found at or below the selected target rather than claiming one quality setting always yields one size.

  3. 3

    Resize only when required

    If quality search cannot meet the limit, width and height are reduced together and the search repeats. A smaller pixel grid often saves more bytes than another small quality reduction, but it also removes spatial detail.

KB means 1,024 bytes here

Choose a 20 KB, 50 KB, 100 KB, 200 KB, 500 KB, or 1 MB target

Each target also has its own page that opens preset: 20 KB, 50 KB, 100 KB, 200 KB, 500 KB, or 1 MB.

Each preset is an upper bound. A successful output can be smaller; an impossible target produces an explicit unmet-target result rather than a false success.

TargetTypical reasonTrade-off to inspect
20 KB / 50 KBTiny thumbnails, strict form limits, or compact previews.Expect aggressive quality reduction, smaller dimensions, or an honest failure for detailed images.
100 KB / 200 KBArticle images, profile pictures, and lightweight product previews.Inspect text edges, faces, gradients, and the final pixel dimensions.
500 KBLarger web content where detail matters more than a tiny transfer.Compare the selected format with the original instead of assuming recompression always helps.
1 MBHigh-resolution sharing or a permissive upload ceiling.An already smaller original-format file is returned unchanged.

Different encoders, different controls

JPEG, WebP, PNG, and Original format do not compress the same way

JPEG and WebP expose a browser encoder quality argument, so the tool can search candidate quality levels. That value is not a universal percentage of visual fidelity and does not predict output size: two photographs at the same dimensions and quality can have very different byte counts because texture, noise, color, and edges compress differently.

Canvas PNG output is lossless for the pixels supplied to the encoder and does not provide an equivalent lossy quality dial. The compressor re-encodes PNG and can reduce its dimensions when the selected target requires fewer pixels. It does not move a decorative quality slider and describe the unchanged PNG behavior as successful compression.

Original format means that the output MIME type matches the validated input type. If its bytes already fit, they are returned unchanged. Once decoding, resizing, or re-encoding occurs, metadata and exact binary identity are not preserved even when the visible format name remains the same.

JPEG
Lossy, no alpha channel, often efficient for photographs.
WebP
Browser-encoded lossy output with alpha support when the browser encoder provides it.
PNG
Lossless pixel encoding; target reduction may require fewer dimensions.
Original format
Keep the validated source format; unchanged bytes only when it already meets the target.

Bounded browser work

Supported input, hard limits, and target failures

CheckContractWhy it matters
File signatureJPEG, PNG, or WebP magic bytes must match a supported static format.A filename or declared MIME type can be missing, mistaken, or renamed.
Input sizeAt most 25 MiB before decoding.Decoded pixels and multiple encoded candidates can consume much more memory than the source file.
Pixel countAt most 40 megapixels, measured as width × height.A small compressed file can still expand to a very large canvas.
UnsupportedHEIC, AVIF, animated PNG/WebP, and other formats are rejected.The tool does not silently keep one animation frame or rely on inconsistent optional decoders.
Encoder MIMEThe browser must return the requested JPEG, PNG, or WebP type.A silent fallback to another format is treated as an error.

Read the result, not the slider

Compression decisions for photographs, screenshots, and graphics

Photograph for a form

Try JPEG or WebP at the form's stated ceiling. Zoom into faces, hair, foliage, and low-light noise; those regions reveal quality loss sooner than a small whole-image preview.

Screenshot with text

Compare PNG and WebP. A tiny JPEG target can introduce halos around high-contrast letters, while PNG may need a dimension reduction to cross the same byte threshold.

Transparent graphic

Use PNG or WebP when alpha matters. JPEG has no transparency, so conversion requires flattening against a chosen background and can permanently change edge pixels.

Already optimized asset

Do not assume another encoding pass helps. If Original format is below the target, preserving its bytes avoids generational loss and needless metadata changes.

Local conversion with explicit boundaries

Image bytes stay in the tab during compression

The selected file is read through browser file APIs, decoded to pixels, encoded into candidate blobs, previewed, and downloaded from the current tab. LiveParse does not intentionally post those file bytes to a remote compression endpoint. Processing can still consume substantial local memory because decoded RGBA pixels use roughly four bytes per pixel before browser overhead and encoded candidates.

This statement is not a universal “zero upload” guarantee for the device. Opening the page makes ordinary website requests; extensions, security software, cloud-synced download folders, clipboard utilities, and any service where you later upload the result have their own behavior. Remove sensitive material or use a controlled offline workflow when those boundaries matter.

Re-encoding is not sanitization. Although canvas output normally does not copy source EXIF, IPTC, or XMP metadata, do not rely on this utility to remove every hidden or sensitive property. Inspect the downloaded file with a dedicated metadata tool before publishing consequential images.

Need dimensions instead of a byte target?

Use the resizer when width, height, percentage, contain, or cover geometry is the primary requirement.

Open the Image Resizer →

Questions answered precisely

Image compressor FAQ

How does the image compressor target a file size?

For JPEG and WebP output, it searches browser-encoder quality values and keeps the highest-quality result found at or below the selected byte target. If quality changes are insufficient, it can reduce width and height proportionally and try again.

Will the compressed image be exactly 100 KB or another selected size?

No. The target is a maximum, not a promise of an exact byte count. Browser encoders produce discrete results, and some images cannot reach a very small target within the tool's quality and dimension safeguards. The result reports whether the target was met.

Does the quality control compress PNG files?

No. The browser Canvas PNG encoder does not expose a lossy quality control. PNG output is re-encoded and may be resized proportionally when needed, but a JPEG-style quality search is not presented as PNG compression.

What happens when the original image is already below the target?

When Original format is selected and the source is already at or below the target, the original file bytes are returned unchanged. Choosing another output format still requires decoding and re-encoding.

Which images can this compressor process?

It accepts validated static JPEG, PNG, and WebP files up to 25 MiB and 40 megapixels. HEIC, AVIF, animated images, files with unsupported signatures, and images the browser cannot decode are rejected.

Can I compress multiple images at once?

Yes. Add up to 20 validated static images with a 100 MiB aggregate limit, remove individual files before processing, then download selected results or one ZIP. Every image stays on your device.

Is my image uploaded while it is compressed?

The selected image is decoded, processed, previewed, and exported in the current browser tab without intentionally sending its bytes to an image-conversion service. Page requests, browser extensions, operating-system services, downloaded files, and destinations where you later share the result remain separate boundaries.