How we test, and what we do not claim
Every number on this site comes from somewhere specific. This page says where, and what it does not cover.
Where the measurements come from
The benchmark table on the HEIC to JPG page was produced by running this site’s own converter — the same code visitors download — on six 12 MP photos (4032 × 3024, 7.1 MB of HEIC in total) in Chrome on an Apple Silicon Mac. Nothing was uploaded; the run happened in a normal tab.
| What we recorded | How | What it means |
|---|---|---|
| Output size | Downloaded file sizes, summed across the six photos | A real figure for this sample at this quality setting |
| Batch time | Wall clock from drop to the last file finishing | Decode and encode included, no upload wait |
| Per-photo cost | Total time divided by six, rounded | About 0.25 s per 12 MP photo on this machine |
| Resized output | Same batch at max width 2048 px | 3.1 MB total, same elapsed time |
Time scales with the device, not with the file count alone. An older laptop takes two to four times longer; a phone takes longer still and is limited by memory before it is limited by speed. We quote the desktop figure and say so.
What we do not claim
- Upload limits are user-reported, not contracts. Figures such as “Gmail 25 MB” or “Instagram 8 MB” are the caps people hit most often. Platforms change them without notice, and every tool page that lists them says as much.
- Round-trip quality is a perceptual judgement. When we say quality 92 is “visually identical”, we mean at normal viewing size on a calibrated-ish display, not that the file is bit-identical. It is not, and it cannot be — JPEG is lossy.
- Facts about other services are dated. The comparison with iLoveIMG was checked against their public pages in November 2026. If they change their free tier, that page is wrong until we re-read it.
- We do not benchmark phones. Mobile browser memory limits are real and vary by model and OS version. Our batch-size guidance is a range, not a guarantee.
How pages get reviewed
Every converter page carries a “last reviewed” month. A review means: re-reading the copy against the current code, re-running one batch to confirm the numbers still hold, checking that every outbound link still resolves, and correcting anything that drifted. Reviews are recorded in the site’s git history, so a change to a claim is a dated, inspectable commit rather than a silent edit.
Sources
Guides cite sources inline and list them at the bottom, with one rule: the link has to load. Where an authoritative page blocks automated access — the ISO catalogue and some vendor portals do — we describe the standard by name instead of linking to a page that may return an error for a reader.
Corrections
If something here is wrong, say so: [email protected]. Factual corrections are made and the page’s updated date changes; the article schema records the new date too. We would rather be corrected than be quoted inaccurately.
Who maintains this
FormatWave is a small project: the people who write the pages are the people who build and run the converter. There are no bylines because there are no separate contributors to credit — and no invented author personas, which is the more common failure mode on sites like this one.