Image Compressor

Compress JPG, PNG, WebP, and GIF images with bounded server-side GD processing. Preview the result, then download it before the temporary file expires.

Temporary server processing
🖼️

Click to upload or drag and drop an image

JPG, PNG, WebP, GIF — up to 10 MB and 12.5 megapixels (4032 × 3024)

Temporary server processing: the selected image is uploaded to this server for GD compression, stored under a random session-bound name, and removed automatically within 5 minutes. Maximum input: 10 MiB and 12.5 megapixels. Downloads use an expiring HMAC-protected URL.
Lower = smaller file. This is not a direct percentage of size reduction.
Server-processing evidence

What this compressor measures and what it does not promise

Image Compressor is the deliberate exception to Z Tools' browser-local model. The selected file is uploaded to PHP, validated against size/dimension limits, decoded and re-encoded with the host's GD codecs, then exposed through a short-lived session-bound download flow. Temporary files are cleaned within 5 minutes.

Current deterministic fixture run

The committed run used PHP 8.3.6 with GD plus JPEG, PNG, and WebP support on a 1200×800 generated gradient/noise fixture.

SettingBaselineOutputSaved
JPEG quality 751,606,187 B152,036 B90.53%
PNG compression 92,885,501 B16,118 B99.44%
WebP quality 75295,832 B116,624 B60.58%

These are synthetic-fixture results, not universal savings claims. Z Tools Lab publishes every tested setting and the codec environment.

Resource and privacy boundaries

  • Maximum upload: 10 MiB, 12.5 megapixels, and 20,000 pixels on either edge. Both file bytes and decoded pixel dimensions matter to server memory use.
  • Temporary storage: filenames are randomized and session-bound; download requests use an expiring HMAC-protected flow rather than serving the upload directory directly.
  • Format behavior: JPEG/WebP quality controls are lossy encoder settings, while PNG uses its own compression level. A numeric quality value is not a perceptual percentage.
  • Already-optimized input: re-encoding can produce a larger file. The tool keeps the original when the generated result is not smaller.
  • No public stress test: benchmark scripts remain server-blocked and are run offline so visitors cannot turn image testing into a CPU/memory endpoint.

For the full workflow, limitations, and why representative real-photo fixtures should be added before making broad compression claims, read Image Compression Decisions.