Online Image Shrinker
OnlineImageShrinker
Back to Learn
Guide

What Platforms Do to Your Images (and How to Upload So They Survive)

You export a crisp image, upload it, and what appears on the feed looks softer, blotchier, and slightly wrong. Colours in the sky have banded into visible steps. Text along the bottom has grown fuzzy edges. Nothing was broken — the platform simply re-encoded your image, and the version everyone sees is not the file you sent.

This happens everywhere, on every platform, and it is not going to stop: serving billions of images means aggressive optimisation is a hard operational requirement. But the process is predictable, and once you know what it does you can upload files that survive it in far better shape.

What actually happens after you hit upload

Broadly, every platform runs your file through the same pipeline:

  1. It gets resized. Every service has maximum dimensions it will store and serve. Upload something wider and it is scaled down — using the platform's own algorithm, tuned for speed at enormous scale rather than for quality on your particular image.
  2. It gets re-compressed, lossily, at the platform's chosen quality. This is the step that does the visible damage. Your carefully-exported quality 90 JPEG is decoded and re-encoded at whatever setting the platform uses — often materially lower.
  3. It gets converted, frequently to WebP or AVIF, and served in multiple sizes to different devices.
  4. Metadata is usually stripped, including EXIF, GPS, and often the colour profile.

Step 2 is the one to focus on, because it is where generation loss enters. Lossy compression discards information permanently. Compressing an image that has already been compressed means quantizing data that has already been quantized — errors from the first pass get treated as real detail and preserved, while genuine detail continues to be discarded. The mechanism is covered in how image compression actually works; the practical consequence is the entire subject of this guide.

Which is why the most important rule is counter-intuitive: do not pre-compress hard. People often shrink an image aggressively before uploading, assuming a smaller file will be treated more gently. The opposite happens — you have spent your quality budget, and the platform then spends it again on top. Upload a high-quality file at the correct dimensions and let the platform do its compression once, from good source material.

Upload at the right dimensions

The single most useful thing you can control is pixel dimensions. Match what the platform actually serves and its resizing step becomes a no-op or close to it.

Common targets, as a practical starting point — platforms adjust these periodically, so treat them as a guide and verify anything critical:

UseDimensionsRatio
Instagram square1080×10801:1
Instagram portrait1080×13504:5
Instagram story / Reel1080×19209:16
YouTube thumbnail1280×72016:9
Pinterest pin1000×15002:3
LinkedIn shared image1200×627~1.91:1
Facebook shared link1200×630~1.91:1
Open Graph / link preview1200×630~1.91:1
Website hero (full-width)1920–2400 widevaries

Two things go wrong here more than anything else.

Uploading something enormous. A 6000 px camera export handed to a platform that serves 1080 px gets scaled down by its algorithm, optimised for throughput. Resizing it yourself first, with a good filter, usually produces a visibly better result — and gives the platform less reason to be aggressive.

Uploading the wrong aspect ratio. If the ratio does not match, the platform either crops it — often centred, cheerfully cutting the top of someone's head off — or letterboxes it. Crop deliberately to the target ratio yourself so you choose what stays in frame. Full details on the up-to-date sizes: social media image sizes.

Give the compressor an easy job

Beyond dimensions, some images survive re-encoding much better than others. The differences follow directly from how lossy compression works, and you can design around them.

Smooth gradients are the most fragile thing you can upload. A clean sky, a subtle background fade, or a soft vignette contains slowly-changing colour across a wide area — exactly what coarse quantization turns into visible banding, those ugly stepped stripes. If a design depends on a gradient, expect the platform to damage it. A very slight amount of noise or grain across the gradient genuinely helps: it gives the encoder irregularity to work with and masks the steps.

Fine text and sharp edges get fuzzy. Lossy compression smears hard transitions — the "ringing" halo around high-contrast boundaries. Small text on a busy background is the worst case. Use larger, heavier type than you think you need, keep contrast high, and avoid thin fonts at small sizes. This is also why a screenshot posted to social media so often looks terrible: it is full of exactly the sharp text edges that lossy encoders handle worst.

Heavy saturation shifts. Very saturated reds in particular tend to smear, because colour information is stored at lower resolution than brightness (chroma subsampling). Pulling extreme saturation back slightly often survives better than it looks like it should.

Busy detail hides damage. The flip side: a photo dense with texture — foliage, fabric, a crowd — absorbs compression artefacts almost invisibly, because there is no smooth area for the errors to show against.

Format, colour, and metadata

Format on upload. JPEG for photographs, PNG for graphics with sharp edges or transparency. Nearly all platforms accept both. WebP support is now common but not universal, and since the platform will convert to its own format anyway, uploading a high-quality JPEG or PNG is the safest choice. On websites you control, that logic reverses — there you should serve WebP or AVIF directly, as covered in optimising images for web speed.

HEIC will trip you up. iPhones shoot HEIC by default and many upload forms reject it outright. Convert first: HEIC to JPG.

Use sRGB. If you export in a wide-gamut profile like Adobe RGB or Display P3 and the platform strips the colour profile — which many do — the raw numbers get reinterpreted as sRGB and your colours shift, usually toward flat and desaturated. This is the real cause of most "my image looks washed out after uploading" complaints. Export sRGB and the problem disappears.

Transparency does not survive JPEG. A logo with a transparent background saved as JPEG gets a solid white or black box behind it. Keep it PNG. See making an image transparent.

Metadata is stripped — do not rely on it, and be glad. Copyright and author fields usually vanish, so do not depend on them for attribution. The upside is a privacy one: EXIF often contains GPS coordinates, exact timestamps, and device details. Most major platforms remove it, but not all services do, and images shared over email, messaging, or direct file transfer frequently retain everything. Stripping it yourself before sharing is the reliable option — how to remove EXIF data.

Messaging apps are the harshest of all

WhatsApp, Messenger, and similar apps compress far more aggressively than social feeds, because they optimise for delivery over slow mobile connections. An image sent in a chat can come out dramatically smaller and visibly degraded, and a photo forwarded repeatedly through several chats degrades further with every hop — each one a fresh round of lossy re-encoding on an already-damaged file.

When quality matters in a chat, use the "send as document" or "send as file" option, which most messaging apps offer. It transfers the original bytes without re-encoding. Sizing guidance for chat: WhatsApp image sizes.

Email has the opposite problem — not quality but size limits, typically around 20–25 MB per message including all attachments. That is a compression target, not a quality one: compressing images for email.

A practical workflow

For anything you care about:

  1. Start from the highest-quality original you have. Never from a copy that has already been through a platform.
  2. Crop to the target aspect ratio deliberately, so you decide what is in frame rather than an automatic centre-crop.
  3. Resize to the platform's serving dimensions — 1080 px wide for most Instagram formats, 1200×630 for link previews, and so on.
  4. Export as JPEG at quality 85–92 in sRGB, or PNG for graphics and transparency. Do not squeeze it harder; leave the platform something to work with.
  5. Strip metadata if the image is personal or was taken at home.
  6. Keep the master file so the next platform, at different dimensions, starts from full quality rather than from a re-compressed copy.

Every one of those steps runs inside your browser here — the image is never uploaded to a server to be processed, which matters for client work and personal photos alike. See in-browser image tools and privacy.

Profile pictures and avatars: the small-size problem

Avatars are the hardest case on any platform, because they combine every failure mode at once. A profile picture is stored small, displayed even smaller, often cropped to a circle, and re-compressed like everything else.

Three rules make them survive:

Design for the circle, not the square. Most platforms mask avatars into a circle at display time while accepting a square upload. Anything in the corners is invisible, and anything near the edge is at risk. Keep the subject centred with generous margin.

Assume it will be seen at 40 pixels. Your avatar appears next to comments and messages at a fraction of its stored size. Fine detail, a full-body photo, or a wordmark with small type turns into unreadable mush. A tightly-cropped face or a single bold symbol is the only thing that reads at that scale.

Upload larger than the display size, but not enormous. Something in the 400–800 px range gives the platform good source material for every size it generates, without handing it a 4000 px file to crush. Profile picture maker.

Thumbnails follow the same logic for the same reason. A YouTube thumbnail is uploaded at 1280×720 but browsed at perhaps 200 px wide on a phone, which is why effective thumbnails use very large text, high contrast, and a single clear focal point rather than a detailed scene.

Troubleshooting: what went wrong

"My colours look washed out." Almost always a colour-profile problem. You exported in Adobe RGB or Display P3, the platform stripped the profile, and the numbers were reinterpreted as sRGB. Re-export in sRGB.

"The sky has stripes in it." Banding from re-compression of a smooth gradient. Add a very slight amount of noise across the gradient before uploading, or accept it — this is the artefact platforms produce most reliably.

"My text went fuzzy." Lossy compression smearing sharp edges. Use larger, heavier type with high contrast, and avoid thin fonts at small sizes.

"The top of my head is cut off." Aspect ratio mismatch triggering an automatic centre-crop. Crop to the platform's ratio yourself before uploading.

"My logo has a white box behind it." It was saved as JPEG, which has no transparency. Use PNG.

"It looks fine on desktop but bad on mobile." The platform serves a smaller, more heavily compressed variant to phones. Check your work at the size it will actually be viewed, not at full screen.

"Quality drops every time I share it onward." Generation loss compounding. Go back to your master file and share a fresh export rather than re-sharing a received copy.

Alt text: the part everyone skips

Every major platform offers a field for image descriptions, and most people leave it empty. It is worth filling in for three separate reasons, only one of which is obvious.

Accessibility. Screen-reader users get nothing from an image without a description. On a platform where images carry the message, an empty alt field means the post simply does not exist for part of your audience.

Discovery. Platforms and search engines read alt text as a signal of what an image contains. On Pinterest and in Google Images especially, a real description does measurable work.

Fallback. When an image fails to load — slow connection, blocked request, an email client refusing remote content — the alt text is what appears in its place.

Writing it well is simpler than people expect: describe what is in the image, specifically, in a sentence. "Woman in a yellow raincoat walking a golden retriever along a wet pier" beats "dog photo" and beats keyword-stuffed strings like "dog pet puppy golden retriever cute animal photo". Skip "image of" and "picture of" — the assistive technology already announces that. And if an image is purely decorative and carries no information, an empty alt attribute is the correct answer on a website, rather than a description of nothing.

Frequently asked questions

Why do my images look worse after uploading? Because the platform decodes your file and re-encodes it at its own quality setting, usually after resizing it. What viewers see is never the file you sent.

Should I compress images before uploading to social media? Resize to the platform's serving dimensions, but do not compress aggressively. Uploading an already-squeezed file means two rounds of lossy damage stacked on each other.

What image size is best for Instagram? 1080×1080 for square, 1080×1350 for portrait (4:5), and 1080×1920 for stories and Reels. Matching the aspect ratio matters as much as the pixel count, since a mismatch triggers an automatic crop.

Does uploading remove EXIF and location data? Most major platforms strip metadata, but not all services do, and files sent by email or direct transfer usually keep everything. If privacy matters, strip it yourself rather than relying on the recipient's platform.

Why does WhatsApp make my photos look so bad? Messaging apps compress far harder than social feeds because they optimise for slow mobile connections, and each forward re-compresses an already-damaged file. Use "send as document" to transfer the original bytes untouched.

Should I upload WebP or AVIF to social platforms? No. Support is inconsistent and the platform converts to its own format regardless. Upload high-quality JPEG or PNG. On a website you control, serving WebP or AVIF directly is the right call.

What size should an Open Graph or link preview image be? 1200×630 (about 1.91:1) is the widely-supported standard for Facebook, LinkedIn, and most link unfurlers.

Key takeaways

  • Every platform re-compresses your upload. What viewers see is never the file you sent.
  • Do not pre-compress aggressively. Uploading an already-squeezed file means two rounds of lossy damage; upload high quality at the right size and let the platform compress once.
  • Match the platform's dimensions and aspect ratio — it is the single biggest thing you control, and it prevents automatic centre-cropping.
  • Gradients band, fine text smears, saturated reds shift — these are the predictable casualties. Add slight noise to gradients and use larger, heavier type.
  • Export in sRGB. Wide-gamut profiles get stripped and reinterpreted, which is why uploads so often look washed out.
  • Messaging apps compress hardest; send as a document or file when quality matters, and avoid repeated forwarding.
  • Keep a master original and export a fresh copy per platform, rather than re-uploading re-compressed versions.

We use cookies to measure how this site is used. No advertising cookies are set. By accepting, you allow these uses. See our Privacy Policy and Cookie Policy.