Ninety-kilobyte rules on upgraded legacy portals
Some upgraded legacy portals cap photos at ninety kilobytes instead of one hundred. Treat ninety as exact, hundred-kilobyte exports fail even when banding looks identical at small size. Start from the original export whenever you can; chat-downloaded copies often arrive already soft. Avoid stacked extensions like .jpg.jpg; intake scripts often choke on them. Width and height on the result card update whenever the engine changes dimensions.
Staying under one hundred without assuming rounding
Do not assume validators round up to one hundred. Aim slightly under ninety when prior attempts failed at the borderline and trust measured bytes on the card. Wide heroes and identity crops behave differently under the same ceiling, choose framing for the actual job. Rename the download to a plain label plus one extension before submission. If the result looks too soft, tighten framing before lowering quality again.
Legacy portal hero photos at ninety KB
Many hero photos work on screen when width is reasonable. Very wide originals may need resize before ninety-kilobyte targeting to avoid mushy skies. Outdoor foliage steals encoder budget, crop tighter when the subject is a person or package. Nearby tools cover adjacent size and format rules when this page is not the match. When format or pixel rules sit outside this page, finish those steps on the related tools linked nearby.
Wide heroes and sky banding near ninety kilobytes
Watch banding in gradients at ninety kilobytes, mild downscaling paired with quality search often beats quality alone on panoramic sources. Copy the exact help-text number when switching pages. Neighboring tools look similar but can still fail validators. If the published rule shifts, switch to the sibling tool that matches the new number. Landing under the published maximum usually beats chasing an exact byte count.
Downscaling before ninety-KB compression
Downscale overly wide heroes before ninety-kilobyte compression when preview shows horizontal striping in blue skies or gray backgrounds. Keep naming clean, odd double extensions confuse some intake scripts. Start from the original export whenever you still have it, not a chat-compressed copy. Portals care about clearing the limit more than hitting a precise byte total.
Ninety versus hundred sibling tools
The hundred-kilobyte sibling targets a different validator integer. Pick the page that matches help text, not whichever preview looks sharper in isolation. Avoid screenshotting the result pane; download the generated file for submission. Avoid screenshotting the preview pane, download the generated file instead. Being under the cap is the real pass condition; exact byte matching is uncommon.
Hero upload checklist at ninety kilobytes
Verify bytes on disk, JPEG MIME expectations, and any width limits before publishing the hero to production CMS fields. Before retrying, open the download once on the same device you will use to submit. Start from the original export whenever you still have it, not a chat-compressed copy. Temporary processing is not archival storage, save the download you intend to submit.
How this target works
The encoder tries quality steps, and may reduce pixels if quality alone cannot get under 90KB. The result card shows the generated file bytes, not a guessed number. A pass means the file is at or under 90KB (92160 bytes).
How we check target tools
Representative target-size tools are run on the same photo set used in our research notes. We record generated bytes and whether the ceiling was met. This page uses that engine. Last methodology review: 5 October 2026. See how we test.
Limits you should expect
A tiny, already-light file may not shrink further. A huge, detailed photo may need fewer pixels to stay under the ceiling, which softens fine texture. The page never claims an exact byte count unless the generated file matches that count.
Technical reference
Everyday “JPEG” files follow the JFIF interchange pattern on top of JPEG compression. MDN summarizes the type, MIME value image/jpeg, and common extensions: MDN image file type guide.
Related research
Target ceilings behave differently from quality-first compression. The 100KB experiment is the measured example: 100KB JPG compression test.
Frequently asked questions
Legacy heroes at 90KB?
Often yes at moderate width; very wide originals may need resize first. Batch mode still shows per-file previews. Review each result before submitting a set.
90KB vs 100KB validator?
Not interchangeable on strict systems, match the published number. Open the download once on your submission device before retrying the portal. Re-read format, pixel, and kilobyte lines together, validators often check more than one.
Downscale wide heroes for 90KB?
When preview shows banding, yes, before or during targeting as dimensions change. Re-read format, pixel, and kilobyte lines together, validators often check more than one.
Borderline 90KB rejections?
Aim a little under ninety kilobytes on the next attempt. Open the download once on your submission device before retrying the portal.
90KB batch heroes?
Each upload is judged independently toward ninety kilobytes. JPEG target-size encoding is lossy; the preview exists so you can judge tradeoffs honestly.
Should I recompress a file that is already under 90KB?
Usually no. Recompress when you changed dimensions or format and the new export went back over ninety kilobytes.