Search FreeImageKit

Jump to any private image tool

Benchmark program

Methodology before conclusions.

The benchmark runner uses the same browser image engine as FreeImageKit’s tools and exports reproducible raw data.

What is measured

JPEG, PNG, WebP and AVIF cases record output bytes, wall-clock duration, successful re-decode and final dimensions. Each case gets one unmeasured warm-up followed by three measured sequential trials.

What timing includes

Duration starts before the job enters the worker and ends when the encoded result returns. It includes worker messaging, source decode, canvas transforms and encoding. Re-decode verification happens afterward and is recorded separately.

Fair comparisons

Use a fixed, redistributable corpus covering photos, UI screenshots, line art, gradients and transparent images. Keep source hashes unchanged between devices. Equal nominal quality settings are not equivalent perceptual quality, so quality-target comparisons must be reported separately.

Every result carries context

Raw runs include source hashes, dimensions, browser identity, platform, hardware hints, exact settings, trial counts, failures and known limitations. The runner validates its output against the versioned contract before download.

What we will not claim

A result from one corpus, browser or device is not a universal truth. Unsupported combinations are marked with a reason instead of being hidden or treated as zero.

Run the local benchmark Download JSON schema