A useful performance claim needs conditions. Image Load Lab does not present a single score or export as a universal truth.
The six-condition record
1. Source
Filename or neutral ID, pixel dimensions, format, file bytes, color and transparency notes.
2. Encoder
Tool, version, output format, quality or lossless mode, resize method, and metadata handling.
3. Page context
URL or fixture, CSS slot width, viewport, device pixel ratio, and whether the image is likely LCP.
4. Runtime
Device, operating system, browser and version, cache state, network profile, and test date.
5. Result
Transferred bytes, selected currentSrc, request timing, relevant Web Vitals trace, and visible-defect notes.
6. Decision
What changed, what stayed uncertain, who reviewed it, and how to roll it back.
Repetition and uncertainty
Timing tests should be repeated under the same conditions. We prefer medians or a visible range to a single best run. Visual quality reviews should include more than one representative image when they inform a workflow.
Measured studies and downloadable fixtures
These five studies apply the record above to original fixtures and publish the files needed to inspect or repeat the run. Results are limited to the documented source, encoders, settings, browser context, and test date.
- JPEG, WebP, and AVIF ladder — fifteen encodes, decoded-pixel metrics, identical crops, source file, and a reproducible bundle.
- Measured JPEG quality ladder — five candidates with bytes, PSNR, luminance SSIM, and crop observations.
- PNG and WebP transparency test — one RGBA fixture composited on four backgrounds, with lossless and lossy outputs.
- Responsive candidate set — four physical WebP widths, hashes, exact markup, and a downloadable
currentSrcfixture. - Measured EXIF removal — one controlled JPEG with a synthetic EXIF segment, byte-level verification, identical decoded pixels, and a reproducible bundle.
The downloadable bundles include portable generation scripts, pinned dependency versions, data files, README instructions, and limitations so another reader can reproduce or challenge the results.
Primary sources
Platform behavior links to the relevant browser, web-standards, Google, or WordPress documentation. Our checklists are editorial recommendations and are labeled as such.
Corrections
If a source changes or a repeatable result contradicts a guide, contact us with the URL, environment, and steps. We will review it and record material changes.